Anda di halaman 1dari 664

Using Your Sybex Electronic Book

To realize the full potential of this Sybex electronic book, you must have Adobe Acrobat Reader with
Search installed on your computer. To find out if you have the correct version of Acrobat Reader, click on
the Edit menu—Search should be an option within this menu file. If Search is not an option in the Edit
menu, please exit this application and install Adobe Acrobat Reader with Search from this CD (double-
click rp500enu.exe in the Adobe folder).

Navigate through the book by clicking on the headings that appear in the left panel;
the corresponding page from the book displays in the right panel.

Search To search, click the Search Query button on the toolbar

or choose Edit >Search > Query to open the Search window. In
the Adobe Acrobat Search dialog’s text field, type the text you
want to find and click Search.

Use the Search Next button (Control+U) and Search

Previous button (Control+Y) to go to other matches in
the book. The Search command also has powerful tools for
limiting and expanding the definition of the term you are
searching for. Refer to Acrobat's online Help (Help > Plug-In
Help > Using Acrobat Search) for more information.

Click here to begin using

your Sybex Elect ronic Book!
Server+ TM

Study Guide
Second Edition

Brad Hryhoruk
Diana Bartley
Quentin Docter

San Francisco • London

Copyright ©2002 SYBEX, Inc., Alameda, CA

Associate Publisher: Neil Edde
Acquisitions and Developmental Editor: Elizabeth Hurley
Editor: Linda Stephenson
Production Editor: Liz Burke
Technical Editors: Andre Paree-Huff, Andy Barkl
Graphic Illustrator: Tony Jonick
Electronic Publishing Specialist: Interactive Composition Corporation
Proofreaders: Nancy Riddiough, Dave Nash, Laurie O’Connell, Richard Ganis
Indexer: Lynnzee Elze
CD Coordinator: Dan Mummert
CD Technician: Kevin Ly
Book Designer: Bill Gibson
Cover Designer: Archer Design
Cover Illustrator/Photographer: Natural Selection

Copyright © 2002 SYBEX Inc., 1151 Marina Village Parkway, Alameda, CA 94501. World rights reserved. No part of this
publication may be stored in a retrieval system, transmitted, or reproduced in any way, including but not limited to photo-
copy, photograph, magnetic, or other record, without the prior agreement and written permission of the publisher.

First Edition copyright © 2001 SYBEX Inc.

Library of Congress Card Number: 2001099339

ISBN: 0-7821-4087-4

SYBEX and the SYBEX logo are either registered trademarks or trademarks of SYBEX Inc. in the United States and/or other

Screen reproductions produced with FullShot 99. FullShot 99 © 1991–1999 Inbit Incorporated. All rights reserved.
FullShot is a trademark of Inbit Incorporated.

The CD interface was created using Macromedia Director, COPYRIGHT 1994, 1997–1999 Macromedia Inc. For more
information on Macromedia and Macromedia Director, visit

Internet screen shot(s) using Microsoft Internet Explorer 5.0 reprinted by permission from Microsoft Corporation.
SYBEX is an independent entity from Citrix Systems, Inc., and not affiliated with Citrix Systems, Inc. in any manner. This
publication may be used in assisting students to prepare for a Citrix Certified Administrator Exam. Neither Citrix Systems,
its designated review company, nor SYBEX warrants that use of this publication will ensure passing the relevant exam. Cit-
rix is either a registered trademark or trademark of Citrix Systems, Inc. in the United States and/or other countries.

TRADEMARKS: SYBEX has attempted throughout this book to distinguish proprietary trademarks from descriptive terms
by following the capitalization style used by the manufacturer.

The author and publisher have made their best efforts to prepare this book, and the content is based upon final release soft-
ware whenever possible. Portions of the manuscript may be based upon pre-release versions supplied by software manu-
facturer(s). The author and the publisher make no representation or warranties of any kind with regard to the completeness
or accuracy of the contents herein and accept no liability of any kind including but not limited to performance, merchant-
ability, fitness for any particular purpose, or any losses or damages of any kind caused or alleged to be caused directly or
indirectly from this book.

Manufactured in the United States of America

10 9 8 7 6 5 4 3 2 1

Copyright ©2002 SYBEX, Inc., Alameda, CA

To Our Valued Readers:

Sybex is proud to have served as a cornerstone member of CompTIA’s Server+ Advisory Committee.
Just as CompTIA is committed to establishing measurable standards for certifying individuals who will
support server environments in the future, Sybex is committed to providing those individuals with the
skills needed to meet those standards. By working alongside CompTIA, and in conjunction with other
esteemed members of the Server+ committee, it is our desire to help bridge the knowledge and skills
gap that currently confronts the IT industry.

In the year since its release, the Server+ has gained industry-wide recognition as a solid indicator of
competency in server technologies. Microsoft recently incorporated the Server+ certification into their
new MCSA (Microsoft Certified Systems Associate) program as an elective option when paired with
CompTIA’s A+ certification. Such integration into vendor-specific certification programs is a strong
endorsement for Server+ and bodes well for those who possess it.

Our authors, editors, and technical reviewers have worked hard to ensure that this Server+ Study Guide
is comprehensive, in-depth, and pedagogically sound. We’re confident that this books will meet and
exceed the demanding standards of the certification marketplace and help you, the Server+ exam can-
didate, succeed in your endeavors.

Good luck in pursuit of your Server+ certification!

Neil Edde
Associate Publisher—Certification
Sybex, Inc.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Software License Agreement: Terms and Conditions
The media and/or any online materials accompanying this If you discover a defect in the media during this warranty
book that are available now or in the future contain pro- period, you may obtain a replacement of identical format at
grams and/or text files (the “Software”) to be used in connec- no charge by sending the defective media, postage prepaid,
tion with the book. SYBEX hereby grants to you a license to with proof of purchase to:
use the Software, subject to the terms that follow. Your pur-
chase, acceptance, or use of the Software will constitute your SYBEX Inc.
acceptance of such terms. Product Support Department
1151 Marina Village Parkway
The Software compilation is the property of SYBEX unless
Alameda, CA 94501
otherwise indicated and is protected by copyright to SYBEX
or other copyright owner(s) as indicated in the media files
(the “Owner(s)”). You are hereby granted a single-user
After the 90-day period, you can obtain replacement media
license to use the Software for your personal, noncommercial
of identical format by sending us the defective disk, proof of
use only. You may not reproduce, sell, distribute, publish,
purchase, and a check or money order for $10, payable to
circulate, or commercially exploit the Software, or any por-
tion thereof, without the written consent of SYBEX and the
specific copyright owner(s) of any component software Disclaimer
included on this media. SYBEX makes no warranty or representation, either
In the event that the Software or components include specific expressed or implied, with respect to the Software or its con-
license requirements or end-user agreements, statements of tents, quality, performance, merchantability, or fitness for a
condition, disclaimers, limitations or warranties (“End-User particular purpose. In no event will SYBEX, its distributors,
License”), those End-User Licenses supersede the terms and or dealers be liable to you or any other party for direct, indi-
conditions herein as to that particular Software component. rect, special, incidental, consequential, or other damages
Your purchase, acceptance, or use of the Software will con- arising out of the use of or inability to use the Software or its
stitute your acceptance of such End-User Licenses. contents even if advised of the possibility of such damage. In
By purchase, use or acceptance of the Software you further the event that the Software includes an online update feature,
agree to comply with all export laws and regulations of the SYBEX further disclaims any obligation to provide this fea-
United States as such laws and regulations may exist from ture for any specific duration other than the initial posting.
time to time. The exclusion of implied warranties is not permitted by some
states. Therefore, the above exclusion may not apply to you.
Software Support
This warranty provides you with specific legal rights; there
Components of the supplemental Software and any offers may be other rights that you may have that vary from state to
associated with them may be supported by the specific state. The pricing of the book with the Software by SYBEX
Owner(s) of that material, but they are not supported by reflects the allocation of risk and limitations on liability con-
SYBEX. Information regarding any available support may be tained in this agreement of Terms and Conditions.
obtained from the Owner(s) using the information provided
in the appropriate files or listed elsewhere on the Shareware Distribution
media. This Software may contain various programs that are distrib-
Should the manufacturer(s) or other Owner(s) cease to offer uted as shareware. Copyright laws apply to both shareware
support or decline to honor any offer, SYBEX bears no and ordinary commercial software, and the copyright
responsibility. This notice concerning support for the Soft- Owner(s) retains all rights. If you try a shareware program
ware is provided for your information only. SYBEX is not the and continue using it, you are expected to register it. Individ-
agent or principal of the Owner(s), and SYBEX is in no way ual programs differ on details of trial periods, registration,
responsible for providing any support for the Software, nor is and payment. Please observe the requirements stated in
it liable or responsible for any support provided, or not pro- appropriate files.
vided, by the Owner(s). Copy Protection
Warranty The Software in whole or in part may or may not be copy-
SYBEX warrants the enclosed media to be free of physical protected or encrypted. However, in all cases, reselling or
defects for a period of ninety (90) days after purchase. The redistributing these files without authorization is expressly
Software is not available from SYBEX in any other form or forbidden except as specifically provided for by the Owner(s)
media than that enclosed herein or posted to therein.

Copyright ©2002 SYBEX, Inc., Alameda, CA

This book is dedicated to my motivation, strength, and support.

—Brad Hryhoruk

To Kara and Abbie.

—Quentin Docter

Copyright ©2002 SYBEX, Inc., Alameda, CA

In creating a project of this magnitude there are a great number of peo-
ple involved. I would like to extend my thanks to everyone who helped make
this book a reality. Liz Burke and Elizabeth Hurley have been wonderfully
supportive and patient with me as I worked toward this goal. Special thanks
to Linda Stephenson for her support and positive criticism throughout this
entire process. I would also like to acknowledge Martial Marcoux. Thank
you for your enduring support and friendship over the years and especially
your support and strength in the last few months. I would not have made it
with out you. Finally I would like to mention my close friends Nancy and
Jerret. The two of you have helped me in more ways then I can ever repay.
—Brad Hryhoruk

I would like to thank my wife Kara for her unwavering support and uncon-
ditional love. I would also like to thank the entire Sybex crew. You are all
great to work with, and very under appreciated.
—Quentin Docter

To the talented staff who assisted us in this undertaking: Liz Burke,

Elizabeth Hurley, Andre Paree-Huff, Andy Barkl, Nancy Riddiough, Dave
Nash, Laurie O’Connell, Richard Ganis, Lynnzee Elze, Stacey Loomis, Dan
Mummert, and Kevin Ly. Thank you for all your hard work and dedication
to this project.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Welcome to the Sybex Server+ Certification Study Guide. The pur-
pose of this text is to assist you in preparing to challenge and succeed at the
CompTIA Server+ Exam. This book was created with the understanding that
you have already successfully completed your A+ certification. Several key
elements in the A+ certification are built upon within the Server+ certifica-
tion. On their website ( CompTIA recommends that can-
didates wishing to achieve Server+ certification first obtain their A+. From
my experience with the Server+ Exam, I recommend also obtaining your
Network+ certification prior to attempting the Server+. Concepts dealing
with networking and network protocols are covered very well in the Net-
work+ Study Guide, by David Groth (Sybex, 2001), and appear within the
Server+ Exam questions.
This text contains several key elements that will assist you on this journey
to Server+ certification. Each chapter begins with an objective outline. The
objectives are outlined in detail later in the introduction, but for now you
should know that they are the key elements that you are expected to know
in detail. This will help you focus your studies on the specifics of the objec-
tives at hand. Each chapter in this book prepares you for the detailed
objectives that are covered in the exam.
Throughout each chapter there are also real world scenarios. These
are real experiences: some describe good situations and others serve as a
warning. Either good or bad, they serve as a means of seeing in action the
information learned.
At times in the chapters, warnings will be given. Performing certain tasks
could pose a threat to server or user safety. Other times the warning is used
to draw your attention to a serious matter that requires your complete focus.
Each time that a warning appears, you will see a small bomb icon.

This is an example of how a warning will be presented within a chapter.

Please be sure to read every warning carefully before proceeding with the

Chapters end with several important elements. These include a chapter

summary, exam essentials, key terms, and review questions.
The chapter summary gives a brief explanation of the key essential com-
ponents of each major topic covered in the chapter. This serves as a refresher

Copyright ©2002 SYBEX, Inc., Alameda, CA

xxiv Introduction

to you as well as a means of finding out if you remember the major features
Exam essentials are brief statements (one sentence) that reemphasize the
most important points that you need to be aware of prior to taking the exam.
Each statement is followed by a brief explanation of why this point is essen-
tial. Be sure to know these essentials before proceeding to the next chapter.
Key terms are exactly what the name implies: a collection of important
terms unique to the chapter and exam. They are defined within the context
of the chapter and then sorted into a list at the end of the chapter. You need
to be aware of these terms as well as their meanings in order to successfully
challenge the Server+ Exam.
The most significant feature in our Study Guides is the practice exam.
Each chapter includes 20 review questions at the end. These practice ques-
tions test your comprehension of the information and key details covered in
each chapter. It is imperative that you work through these chapter tests.
They not only help you remember the information presented in the chapters,
but also assist you in preparing for the real Server+ Exam.

Don’t just study the questions and answers—the questions on the actual
exam will be different from the practice ones included in this book and on the
CD. The exam is designed to test your knowledge of a concept or objective, so
use this book to learn the objective behind the question.

What Is Server+ Certification?

T he Server+ certification was created by the Computer Technology
Industry Association (CompTIA) with the purpose of providing a vendor-
neutral means of certifying the competency of a Server Hardware Specialist.
A Server Hardware Specialist is someone who spends time solving problems
to ensure that servers are functional and applications remain available. The
specialist should have an in-depth understanding of how to plan a network
and how to install, configure, and maintain a server. This should include
knowing the hardware that goes into a server implementation, how data
storage subsystems work, the basics of data recovery, and how I/O sub-
systems work. CompTIA recommends that Server+ candidates have between

Copyright ©2002 SYBEX, Inc., Alameda, CA

Introduction xxv

18 and 24 months of experience in the server technology industry, as well as

experience working on a server. This will provide you with the essential
hands-on experience to understand as well as work through the concepts
successfully. A Server+ specialist should be able to demonstrate competency
in the following areas:
 Has in-depth knowledge of servers, including working knowledge of
troubleshooting, physical security, and disaster recovery. Can recover
from a server failure.
 Ensures high availability by meeting the Service Level Agreement
requirements, including proactively recognizing and responding to
problems and performing recovery.
 Has thorough working knowledge of hardware configuration and net-
work connectivity. Includes the ability to perform problem determi-
nation for all aspects of the server (hardware, software, networking).
 Installs and configures server hardware to meet application
 Implements current and emerging data storage and transfer technolo-
gies such as SCSI and RAID.
 Has a thorough working knowledge of networking protocols (e.g.,
TCP/IP, IPX/SPX, SNMP, DMI) for diagnosing the impact of the net-
work on the server and vice versa.
 Provides support, including second-level support, for resellers and end
 Performs maintenance on server systems, data storage subsystems,
and network devices.
 Has good planning and integration skills to be able to upgrade a server
without impacting network users; increase storage capacity without
impacting network users; design and implement a data recovery plan
in the event of a network device failure; and perform peripheral
upgrades, BIOS upgrades, memory upgrades, processor upgrades,
mass storage upgrades, and adapter upgrades
Demonstrates high levels of leadership through mentoring and training
others in server concepts and operations.

Copyright ©2002 SYBEX, Inc., Alameda, CA

xxvi Introduction

Why Become Server+ Certified?

Obtaining the Server+ Certification will open several doors for you.
Not only is it gratifying to know you have obtained a personal growth
achievement, but you are also improving your employability skills. The
Server+ certification demonstrates a proof of professional achievement.
You will receive a certificate for framing and a wallet card. It increases your
marketability in today’s information technology–driven industry. Having
your Server+ certification also will help your pursuit for advancement in
the computer field. If you are hoping to become involved or more involved
in computer technical work, having your Server+ certification will assist
tremendously in validating your commitment and expertise.
There are a number of reasons for becoming Server+ certified:
 It demonstrates proof of professional achievement.
 It increases your marketability.
 It provides greater opportunity for advancement in your field.
 It is increasingly found as a requirement for some types of advanced
 It raises customer confidence in you and your company’s services.
Let’s explore each reason in detail.

Provides Proof of Professional Achievement

Specialized certifications are the best way to stand out from the crowd. In
this age of technology certifications, you will find hundreds of thousands of
administrators who have successfully completed the Microsoft and Novell
certification tracks. To set yourself apart from the crowd, you need a little bit
more. The Server+ exam is the starting point for the Server Specialist and will
give you the recognition you deserve.

Increases Your Marketability

Almost anyone can bluff their way through an interview. Once you have cer-
tified in an area such as Server+, you will have the credentials to prove your
competency. And certifications are not something that can be taken from
you when you change jobs. Once certified, you can take that certification
with you to any of the positions you accept.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Introduction xxvii

Provides Opportunity for Advancement

Those individuals who prove themselves as competent and dedicated are the
ones who will most likely be promoted. Becoming certified is a great way to
prove your skill level and shows your employers that you are committed to
improving your skill set. Look around you at those who are certified. They
are probably the ones who receive good pay, raises, and promotions when
they come up.

Fulfills Training Requirements

Many companies have set training requirements for their staff so that they
stay up-to-date on the latest technologies. Having a Server+ certification pro-
vides administrators another certification path to follow when they have
exhausted some of the other industry-standard certifications.

Raises Customer Confidence

As companies discover the advantages of having staff with the Server+ cer-
tification, they will undoubtedly want staff to challenge for this certification.
Many companies outsource the work to consulting firms with experience
working with servers. Those firms that have certified staff have a definite
advantage over other firms that do not.

How to Become Server+ Certified

There is only one exam that you must pass in order to achieve your
Server+ certification (SKO-001). This exam is administered by Sylvan
Prometric and can be taken at any authorized Prometric testing center. To
locate the closest testing center to you, call 1 800 755-EXAM. If you want
to register for your Server+ Exam, you can call the exam registration at
1-800-776-4276. If you have not booked an exam through Sylvan Prometric
before, you will need give them your name, social security number, mailing
address, phone number, employer, and a credit card number for payment. If
you choose, you can also register for the test at the Sylvan Prometric online
registration site (
In order to obtain your Server+ certification, you must achieve a passing
mark of at least 75 percent on the 80-question exam in the 90 minutes

Copyright ©2002 SYBEX, Inc., Alameda, CA

xxviii Introduction

allocated. All the questions are multiple choice and can contain one or
more answers. The questions are often tricky, often with several possibly
correct answers. You must select the most correct answer. Be sure to read
each question carefully. The Server+ Exam at this time is not adaptive. This
means that you can skip questions and come back to them at a later point in
the exam. CompTIA has not announced a date that the Server+ Exam will
become adaptive, or in fact whether it will at all.

Who Should Buy This Book?

I f you want to acquire a solid foundation in Server+, this book is for
you. You’ll find clear explanations of the concepts you need to grasp and
plenty of help to achieve the high level of professional competency you need
in order to succeed in your chosen field.
If you want to become certified as a Server+ technician, this book is
definitely for you. However, if you just want to attempt to pass the exam
without really understanding servers, this study guide is not for you. It is
written for people who want to acquire hands-on skills and in-depth
knowledge of network servers.

How to Use This Book and the CD

W e’ve included several testing features in both the book and on the
CD-ROM bound at the front of the book. These tools will help you retain
vital exam content as well as prepare to sit for the actual exam. Using our
custom test engine, you can identify weak areas up front and then develop a
solid studying strategy using each of these robust testing features. Our thor-
ough readme will walk you through the quick and easy installation process.
Before You Begin At the beginning of the book (right after this intro-
duction, in fact) is an assessment test that you can use to check your readi-
ness for the actual exam. Take this test before you start reading the book.
It will help you determine the areas you may need to brush up on. The
answers to each assessment test appear on a separate page after the last
question of the test. Each answer also includes an explanation and a note
telling you in which chapter this material appears.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Introduction xxix

Chapter Review Questions To test your knowledge as you progress

through the book, there are review questions at the end of each chapter.
As you finish each chapter, answer the review questions and then check to
see if your answers are right—the correct answers appear on the page fol-
lowing the last review question. You can go back to reread the section that
deals with each question you got wrong to ensure that you get the answer
correctly the next time you are tested on the material.
Electronic “Flashcards” You’ll also find 150 flashcard questions for on-
the-go review. Download them right onto your Palm device for quick and
convenient reviewing.
Test Engine In addition to the assessment test and the chapter review
tests, you’ll find two sample exams. Take these practice exams just as if
you were taking the actual exam (i.e., without any reference material).
When you have finished the first exam, move onto the next one to solidify
your test-taking skills. If you get more than 90 percent of the answers cor-
rect, you’re ready to go ahead and take the certification exam.
Full Text of the Book in PDF Also, if you have to travel but still need to
study for the Server+ exam and you have a laptop with a CD-ROM drive,
you can carry this entire book with you just by taking along the CD-ROM.
The CD-ROM contains this book in PDF (Adobe Acrobat) format so it
can be easily read on any computer.

Exam Objectives
A s with the other CompTIA certifications, a series of exam objectives or
topics have been identified by the Advisory Committee as being key to
becoming certified as a competent technician. In the Server+ certification,
these objectives fall under seven major areas: installation, configuration,
upgrading, proactive maintenance, environment, troubleshooting and prob-
lem determination, and disaster recovery. Each key area is weighted on the

Exam objectives are subject to change at any time without prior notice and
at CopmTIA’s sole discretion. Please visit the Certification page of CompTIA
website at for the most current listing of exam objectives.

Copyright ©2002 SYBEX, Inc., Alameda, CA

xxx Introduction

exam differently. The exam weights are set to focus on the areas that a server
technician needs to be most knowledgeable in.
Behind every computer industry exam you are sure to find exam objec-
tives—the broad topics on which the exam developers want to ensure your
competency. The official Server+ exam objectives are listed here.

Server+ Exam Blueprint

Job Dimension % of Exam (approximate)
1.0 Installation 17%
2.0 Configuration 18%
3.0 Upgrading 12%
4.0 Proactive Maintenance 9%
5.0 Environment 5%
6.0 Troubleshooting and Problem 27%
7.0 Disaster Recovery 12%

1.0 Installation (17%)

1.1 Conduct pre-installation planning activities
 Plan the installation
 Verify the installation plan
 Verify hardware compatibility with operating system
 Verify power sources, space, UPS and network availability
 Verify that all correct components and cables have been delivered
1.2 Install hardware using ESD best practices (boards, drives, processors,
memory, internal cable, etc.)
 Mount the rack installation
 Cut and crimp network cabling
 Install UPS

Copyright ©2002 SYBEX, Inc., Alameda, CA

Introduction xxxi

 Verify SCSI ID configuration and termination

 Install external devices (e.g., keyboards, monitors, subsystems,
modem rack, etc.)
 Verify power-on via power-on sequence

2.0 Configuration (18%)

2.1 Check/upgrade BIOS/firmware levels (system board, RAID, controller,
hard drive, etc.)
2.2 Configure RAID
2.3 Install NOS
 Configure network and verify network connectivity
 Verify network connectivity
2.4 Configure external peripherals (UPS, external drive subsystems, etc.)
2.5 Install NOS updates to design specifications
2.6 Update manufacturer specific drivers
2.7 Install service tools (SNMP, backup software, system monitoring
agents, event logs, etc.)
2.8 Perform Server baseline
2.9 Document the configuration

3.0 Upgrading (12%)

3.1 Perform full backup
 Verify backup
3.2 Add Processors
 On single processor upgrade, verify compatibility
 Verify N 1 stepping
 Verify speed and cache matching
 Perform BIOS upgrade

Copyright ©2002 SYBEX, Inc., Alameda, CA

xxxii Introduction

 Perform OS upgrade to support multiprocessors

 Perform upgrade checklist, including: locate/obtain latest test drivers,
OS updates, software, etc.; review FAQs, instruction, facts and issues;
test and pilot; schedule downtime; implement ESD best practices;
confirm that upgrade has been recognized; review and baseline; doc-
ument upgrade.
3.3 Add hard drives
 Verify that drives are the appropriate type
 Confirm termination and cabling
 For ATA/IDE drives, confirm cabling, master/slave and potential
cross-brand compatibility
 Upgrade mass storage
 Add drives to array
 Replace existing drives
 Integrate into storage solution and make it available to the operating
 Perform upgrade checklist, including: locate and obtain latest test
drivers, OS updates, software, etc.; Review FAQs, instructions, facts
and issues; test and pilot; schedule downtime; implement using ESD
best practices; confirm that the upgrade has been recognized; review
and baseline; document the upgrade.
3.4 Increase memory
 Verify hardware and OS support for capacity increase
 Verify memory is on hardware/vendor compatibility list
 Verify memory compatibility (e.g., speed, brand, capacity, EDO,
 Perform upgrade checklist including: locate and obtain latest test
drivers, OS updates, software, etc.; review FAQs, instructions, facts
and issues; test and pilot; schedule downtime; implement using ESD
best practices; confirm that the upgrade has been recognized; review
and baseline; document the upgrade

Copyright ©2002 SYBEX, Inc., Alameda, CA

Introduction xxxiii

 Verify that server and OS recognize the added memory

 Perform server optimization to make use of additional RAM
3.5 Upgrade BIOS/firmware
 Perform upgrade checklist including: locate and obtain latest test
drivers, OS updates, software, etc.; review FAQs, instructions, facts
and issues; test and pilot; schedule downtime; implement using ESD
best practices; confirm that the upgrade has been recognized; review
and baseline; document the upgrade
3.6 Upgrade adapters (e.g., NICs, SCSI cards, RAID, etc.)
 Perform upgrade checklist including: locate and obtain latest test
drivers, OS updates, software, etc.; review FAQs, instructions, facts
and issues; test and pilot; schedule downtime; implement using ESD
best practices; confirm that the upgrade has been recognized; review
and baseline; document the upgrade
3.7 Upgrade peripheral devices, internal and external
 Verify appropriate system resources (e.g., expansion slots, IRQ,
DMA, etc.)
 Perform upgrade checklist including: locate and obtain latest test
drivers, OS updates, software, etc.; review FAQs, instructions, facts
and issues; test and pilot; schedule downtime; implement using ESD
best practices; confirm that the upgrade has been recognized; review
and baseline; document the upgrade
3.8 Upgrade system-monitoring agents
 Perform upgrade checklist including: locate and obtain latest test
drivers, OS updates, software, etc.; review FAQs, instructions, facts
and issues; test and pilot; schedule downtime; implement using ESD
best practices; confirm that the upgrade has been recognized; review
and baseline; document the upgrade
3.9 Upgrade service tools (e.g., diagnostic tools, EISA configuration,
diagnostic partition, SSU, etc.)
 Perform upgrade checklist including: locate and obtain latest test
drivers, OS updates, software, etc.; review FAQs, instructions, facts
and issues; test and pilot; schedule downtime; implement using ESD

Copyright ©2002 SYBEX, Inc., Alameda, CA

xxxiv Introduction

best practices; confirm that the upgrade has been recognized; review
and baseline; document the upgrade
3.10 Upgrade UPS
 Perform upgrade checklist including: locate and obtain latest test
drivers, OS updates, software, etc.; review FAQs, instructions, facts
and issues; test and pilot; schedule downtime; implement using ESD
best practices; confirm that the upgrade has been recognized; review
and baseline; document the upgrade

4.0 Proactive Maintenance (9%)

4.1 Perform regular backup
4.2 Create baseline and compare performance
4.3 Set SNMP thresholds
4.4 Perform physical housekeeping
4.5 Perform hardware verification
4.6 Establish remote notification

5.0 Environment (5%)

5.1 Recognize and report on physical security issues
 Limit access to server room and backup tapes
 Ensure physical locks exist on doors
 Establish anti-theft devices for hardware (lock server racks)
5.2 Recognize and report on server room environmental issues (tempera-
ture, humidity/ESD/ power surges, back-up generator/fire suppression/
flood considerations)

6.0 Troubleshooting and Problem Determination (27%)

6.1 Perform problem determination
 Use questioning techniques to determine what, how, when.
 Identify contact(s) responsible for problem resolution

Copyright ©2002 SYBEX, Inc., Alameda, CA

Introduction xxxv

 Use senses to observe problem (e.g., smell of smoke, observation of

unhooked cable, etc.)
6.2 Use diagnostic hardware and software tools and utilities
 Identify common diagnostic tools across the following OS: Microsoft
Windows NT/2000; Novell Netware, UNIX, Linux, IBM OS/2
 Perform shut down across the following OS: Microsoft Windows NT/
2000, Novell Netware, UNIX, Linux, IBM OS/2
 Select the appropriate tool
 Use the selected tool effectively
 Replace defective hardware components as appropriate
 Identify defective FRUs and replace with correct part
 Interpret error logs, operating system errors, health logs, and critical
 Use documentation from previous technician successfully
 Locate and effectively use hot tips (e.g., fixes, OS updates, E-support,
web pages, CDs)
 Gather resources to get problem solved: identify situations requiring
call for assistance; acquire appropriate documentation
 Describe how to perform remote troubleshooting for a wake-on-LAN
 Describe how to perform remote troubleshooting for a remote alert.
6.3 Identify bottlenecks (e.g., processor, bus transfer, I/O, disk I/O,
network I/O, memory)
6.4 Identify and correct misconfigurations and/or upgrades
6.5 Determine if problem is hardware, software or virus related

7.0 Disaster Recovery (12%)

7.1 Plan for disaster recovery
 Plan for redundancy (e.g., hard drives, power supplies, fans, NICs,
processors, UPS)

Copyright ©2002 SYBEX, Inc., Alameda, CA

xxxvi Introduction

 Use the technique of hot swap, warm swap and hot spare to ensure
 Use the concepts of fault tolerance/fault recovery to create a disaster
recovery plan
 Develop disaster recovery plan
 Identify types of backup hardware
 Identify types of backup and restoration schemes
 Confirm and use off site storage for backup
 Document and test disaster recovery plan regularly, and update
as needed
7.2 Restoring
 Identify hardware replacements
 Identify hot and cold sites
Implement disaster recovery plan.

Tips for Taking the Server+ Exam

As you reach your final steps in preparing for the Server+ Exam, keep a few
general tips in mind:
Try to arrive at your exam early. This will allow you to familiarize
yourself with the location and exam setting. You can (try to) relax and
collect yourself for the task ahead.
Bring two forms of identification. It helps to have a photo identifica-
tion. You will be required to sign in and the examiner will perform some
last-minute registration with you.
Ask for a pen and sheet of paper. You are allowed to have these items
at your computer station with you. Many people are unaware of this, or
don’t take advantage of it. If you have to refer to charts, tables, or infor-
mation that you crammed in at the last minute, they can be written down
on this paper. When you are sitting at your station, the clock on your
exam does not start until you click Start. You can spend some time writing
your notes at this point. When you have written out all the information
that you need to, then click Start on the computer.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Introduction xxxvii

Read the questions carefully. Numerous questions are written with

multiple correct or apparently correct answers. You must select the most
appropriate answer. Many discussion groups based on certification exams
have complaints and in-depth discussions on specific questions, as they
appear so open-ended.
Don’t leave questions unanswered. You have a better chance of getting
a point if you guess.
Because the test is not adaptive, you can mark questions for review so you
can go back to them later. If you are not positive about your answer,
use this feature. It is possible that questions or answers later in the test will
tip you off to the questions that stumped you.
Questions can have more than one correct answer. In the test, answer
blanks will appear as check boxes, not radio dials. If more than one
option should be selected, the question will inform you that you are to
select multiple answers. If you select too many answers, the testing soft-
ware will prompt you—and this can come in handy if you are unsure of
how many answers are correct. Check all the answers that you feel could
be correct, and if the software alerts you that you have selected too many
responses, you can then eliminate the weakest options.
Work through the test in several phases. Your first run-through should
be to answer the easy questions as well as reading through all the ques-
tions and answers to familiarize yourself with what you are presented
with. Next go through and answer the medium-difficulty questions and
review your easy questions and answers. Save the hard ones for last. This
will let you know exactly how many hard ones there are left and the
amount of time you have remaining to deal with them. This way you can
best budget your time to complete the test.
There are many different websites with information pertaining to certifi-
cation exams. Many of these websites claim to have exam questions and
exam information posted on them. Part of the CompTIA exam require-
ment is that you agree not to share exam information after you have com-
pleted your test. You will ultimately decide on your own whether or not
to visit these websites. Many aspiring technicians feel that it takes away
from the validity of the exams and the certification. Another major con-
cern is the validity of the information presented on these websites. Every-
one appears to be an expert, yet there is a lot of incorrect information

Copyright ©2002 SYBEX, Inc., Alameda, CA

xxxviii Introduction

posted. I would recommend not using these exam information websites

for this reason alone.
Good luck on your path to Server+ certification.

About the Authors

Brad Hryhoruk has been actively instructing for over 10 years. He has
his bachelor of Education degree as well as A+, Network+, Inet+, Server+,
and MCP. He currently teaches at the Aboriginal Community Campus in
Winnipeg Manitoba. Brad also instructs certification programs as well as
working as a project manager. You can reach Brad at
Diana Bartley, B.ED., MCSE, MCP+I, MCT, A+, Network+, i-Net+, and
Server+, is currently an MCSE instructor and technical writer for various
IT-related topics. She has co-authored many books covering Windows 2000
Directory Services, MCSE Clustering, and ISA Server.
Quentin Docter, MCSE, MCT, CCNA, CNE, A+ and Server+, is a 9-year
industry veteran with experiences ranging from administrator, consultant, to
instructor. He is currently working as an independent consultant and author.
Quentin has worked on 7 books for Sybex, most recently, the MCSE
Windows 2000 Network Infrastructure Design Study Guide.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment Test
1. Name three ways adapters can work together (select all that apply).

A. Adapter grouping

B. Adapter fault tolerance

C. Adapter virtual private networks

D. Adapter Load Balancing

E. Adapter teaming

2. What are three possible configurations for an ATA/IDE device (select

all that apply)?
A. Master, with slave present
B. Slave, with a master present

C. Slave, no master present

D. Master, no slave present

3. If you have a RAID 3 system made up of four 20GB drives, how much
usable disk storage space would you have?
A. 80GB

B. 60GB

C. 40GB
D. 20GB

4. You want to filter packets of certain TCP/IP types coming in from and
going out to the Internet. What type of server application do you
A. Firewall

B. Proxy server

C. Router

D. Gateway

Copyright ©2002 SYBEX, Inc., Alameda, CA

xl Assessment Test

5. Pick the levels of cache that can be present in a computer with a

Pentium III Gigahertz processor.
A. L1

B. L2

C. L3

D. L4

6. SNMP is part of what protocol suite?




D. AppleTalk

7. What type of server resolves DNS names to IP addresses?





8. What is the default ID setting for a SCSI host bus adapter?

A. 7

B. 5
C. 3

D. 6

9. Which of the following TCP/IP addresses is in a private address range?





Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment Test xli

10. With which Internet standard protocol is Active Directory accessed?





11. A BNC connector is used on what type of Ethernet implementation?

A. Thinnet

B. Thicknet



12. In every SCSI-3 bus, how many terminators are there?

A. Four

B. Three

C. Two

D. One

E. One per device

13. Your server’s backups are taking so long that you find you must start
them as you’re leaving work for the day and they often don’t finish
until noon the next day. What are some options you can consider
(select all that apply)?
A. Set up differential backups.

B. Add more backup tape drives.

C. Cut down on the number of files being backed up.

D. Set up incremental backups.

14. Which of the following is a memory error check?




Copyright ©2002 SYBEX, Inc., Alameda, CA
xlii Assessment Test

15. You want to set up your computer room so that an exact duplicate
exists in a different building and you somehow replicate the informa-
tion on the servers in your production room to this duplicate room.
What kind of function are you performing?
A. Backups

B. Fault tolerance

C. High-availability

D. Disaster recovery

16. Single mode fiber optics uses which of the following as a light source?

A. Laser

C. Fluorescent

D. Incandescent

17. Memory Interleaving is another way of doing which of the following:

A. Error checking

B. Accessing information stored on the memory chip

C. Determining parity

D. Installing chips

18. What are some common diagnostic tools that you can utilize no
matter what NOS you’re working with (choose all that apply)?
A. Event logs

B. TCP/IP software


D. BIOS utilities

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment Test xliii

19. How is a PCI bus configured?

A. Through jumpers

B. On the motherboard

C. Through the system BIOS

D. With a CD-ROM

20. Which of the following is a server-specific CPU?

A. Intel Itanium

B. AMD Duron

C. Intel Celeron

D. AMD Athlon

21. Name some areas of concern to look at when you are attempting to
diagnose system bottlenecks (select all that apply).
A. IRQ conflicts

B. CPU speed

C. Hard disk RPMs

D. SCSI version

22. What kind of hard disks will typically be installed in a RAID 5 system?




23. Most servers today are equipped with what kind of system memory




Copyright ©2002 SYBEX, Inc., Alameda, CA

xliv Assessment Test

24. Why are maintenance logs important?

A. They provide a clear picture of what the service techs have been
B. They provide a background of what has been done to a computer.

C. They provide an instruction manual for doing routine tasks.

25. When cabling a building, what should you do?

A. Only use fiber optic cable.

B. Always use copper conduit.

C. Always check local building codes.

D. Assume that you do not need a permit.

26. How many terminators are there on a Thinnet network?

A. One

B. Two

C. One for every 50 hosts

D. One for every 100 hosts

27. Which is true of fiber optics?

A. It is affected by EMI.

B. It is affected by heat.

C. The cable can be made of glass.

D. The cable is always made of copper.

28. You have a single network card with four ports on it. What can that
card not be configured to do?
A. Adapter Load Balancing

B. Adapter teaming

C. Adapter fault tolerance

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment Test xlv

29. Four network cards grouped together for Load Balancing will have
how many IP addresses?
A. Four

B. Three

C. Two

D. One

30. What category of UTP is rated for 1000Mbps transfer speeds?

A. Category 3

B. Category 5

C. Category 4

D. All of the above

31. You have just purchased a motherboard that supports dual proces-
sors. Which Pentium III processors can be used on the board?
A. Any Xeon with any P-II

B. P-IIIs of the same speed

C. Any P-III

D. Any P-II with any P-III

32. What happens when a parity-checking memory module determines

that corruption has occurred?
A. The problem is immediately corrected and the end user is none
the wiser.
B. An error message pops up on the screen describing the error to the
end user and giving the user a chance to fix the problem.
C. An entry is made in the memory error log, but the system continues
to operate.
D. The system is halted.

Copyright ©2002 SYBEX, Inc., Alameda, CA

xlvi Assessment Test

33. What does ESD stand for?

A. Electromagnetic static discharge.

B. Electronic static device.

C. Environmental static discharge.

D. Electrostatic discharge.

34. When carrying memory chips from one place to another, what type of
ESD equipment should you use?
A. Wrist strap.

B. ESD vest.

C. Antistatic bag.

D. No ESD protection is required.

35. What is the plenum?

A. The type of metallic shielding surrounding a fiber optic cable

B. The type of cable used in fiber optic installations

C. The air space between the ceiling and the actual roof of a building

D. Precious metal like gold

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Assessment Test xlvii

Answers to Assessment Test

1. B, D, E. Adapters can work together with Load Balancing, fault
tolerance, or teaming. See Chapter 6 for further information.
2. A, B, D. IDE devices can be a master with no slave present, a master
with a slave present, and a slave with a master present. See Chapter 4
for more information.
3. B. You would have three 20GB drives for storage and one 20GB
drive for parity. Therefore, you would have 60GB of usable storage
space. See Chapter 4 for more information.
4. A, B. Proxy servers and firewalls are closely comparable. Firewalls
perform network address translation (NAT), taking a private net-
work’s IP address and converting it to a public address; they also filter
incoming and outgoing packets. A proxy server filters incoming and
outgoing packets and might be able to NAT as well, though you prob-
ably didn’t buy proxy server software for its network address transla-
tion capabilities. Both firewalls and routers can prevent packets of
certain types from going out to the Internet or coming in from the
Internet. Routers route packets, and gateways provide an opening
to a different system or environment. See Chapter 1 for more
5. A, B. Level 1 and Level 2 cache can only be present with a Pentium III
processor. See Chapter 3 for further information.
6. A. Simple Network Management Protocol is part of the TCP/IP
protocol suite. See Chapter 12 for more information.
7. B. A DNS Server resolves a DNS name to an IP address. See Chapter 1
for further information.
8. A. The SCSI bus adapter uses ID 7 by default. See Chapter 4 form
further information.
9. D. There are three IP address ranges that are considered to be private:,, and See Chapter 8 for further
10. C. Active Directory is based on LDAP. See Chapter 1 for further

Copyright ©2002 SYBEX, Inc., Alameda, CA

xlviii Answers to Assessment Test

11. A. BNC connectors are used on Thinnet Ethernet networks. See

Chapter 6 for further information.
12. C. There will always be two terminators, one at each end of the SCSI
chain. See Chapter 4 for further information.
13. B, C, D. You can add backup tape drives to the system, thus provid-
ing more places for backups. You can also groom the file list to see if
there are files that are being backed up that don’t actually need to be.
You can also go to a system where you perform a full backup over the
weekend and then use incremental backups during the week. Incre-
mental backups only back up the files that have changed since the
previous day’s backup. See Chapter 14 for further information.
14. B. ECC stands for error-correcting code and is used as a memory
error checker in RAM. See Chapter 12 for further information.
15. D. When you have a room in a different building that contains dupli-
cate computing gear and also contains data that has been replicated to it
from the primary servers, you’re working with a disaster recovery (DR)
methodology. DR is, as you might imagine, expensive. The technology
has grown and become viable for entities needing assurance that, should
something disastrous happen, they could continue on with everyday
business functions. See Chapter 15 for further information.
16. A. Single mode fiber optics technology uses a laser as the light source.
This is the more expensive form of fiber optic transmissions, but the
most efficient. See Chapter 6 for further information.
17. B. Memory interleaving is a way of quickly getting access to infor-
mation stored on the memory chip. See Chapter 12 for further
18. A, B, D. All network operating systems create logs that inform you
of critical events. The method you use for reading the logs might be
different, but you can be assured that event logs are generated within
any NOS. You can use TCP/IP software commands such as PING and
NSLOOKUP to perform basic network connectivity troubleshooting.
You can use the server BIOS utilities and peripheral utilities to check to
see if the hardware is correctly configured. Option C would not be an
answer because not all machines have DOS installed on them. See
Chapter 12 for further information.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Assessment Test xlix

19. C. A PCI bus is configured through the system BIOS. See Chapter 3
for further information.
20. A. The Intel Itanium is a server-specific processor. All others listed
were originally created for a desktop computer. See Chapter 3 for fur-
ther information.
21. B, C, D. A CPU’s speed can create a bottleneck if the applications
and users trying to access the computer outpace the speed with which
the processor can answer requests. In SCSI you’ll also need to be con-
cerned about the hard disk’s RPMs and the SCSI version level. See
Chapter 12 for further information.
22. C. In almost all cases (except systems that utilize software RAID)
you’ll use SCSI drives in your RAID array. See Chapter 4 for further
23. D. Most commercial servers today come equipped with 72-pin Dual
Inline Memory Modules (DIMMs). The reason for this is twofold: you
get 64-bit memory and SIMMs need to be installed in pairs whereas
DIMMs can be installed singly. See Chapter 3 for further information.
24. B. Maintenance logs provide a background of what has been done to
a computer. See Chapter 9 for further information.
25. C. When cabling a building, you should check the local building
codes. These codes will vary by locality. See Chapter 6 for further
26. B. There is a 50-ohm terminator at each end of the bus on a Thinnet
network. See Chapter 6 for further information.
27. C. Fiber cable is made of glass or plastic. See Chapter 6 for further
28. C. Fault tolerance requires more than one card, not more than one
port. See Chapter 6 for further information.
29. D. A group of network cards used in Load Balancing will have one IP
address. See Chapter 6 for further information.
30. B. Category 5 is the only option that supports transfer speeds of
1000Mbps. See Chapter 6 for further information.

Copyright ©2002 SYBEX, Inc., Alameda, CA

l Answers to Assessment Test

31. B. With Pentium III processors, the multiplier and the FSB must
match. See chapter 3 for further information.
32. D. With parity, if it is determined that there has been some corruption,
the system is halted. See Chapter 12 for further information.
33. D. ESD is electrostatic discharge—rapid discharge of static electricity
from one conductor to another of a different potential. If your body is
holding a static charge and you touch an electronic component, that
discharge can seriously damage the electronics. See Chapter 15 for
further information.
34. C. Antistatic bags are used to carry electronic equipment from one
place to another. This will prevent ESD from damaging the chips. See
Chapter 15 for further information.
35. C. The plenum is the space created for air circulation between a
drop-down ceiling and the roof, or under a raised floor; this space is
commonly used to run cables. See Chapter 6 for further information.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Installation PART

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Server Types and Roles

Copyright ©2002 SYBEX, Inc., Alameda, CA

F or the most part, each chapter in this book closely maps to
the Server+ Exam Blueprint from CompTIA. We have listed the exam objec-
tives covered by each chapter at the start of the chapter, and fully covered
each objective to the best of our ability in that one chapter. In a couple of
cases objectives cross chapter lines, but this is infrequent, and is done to
arrange the information in the most readable and useful format possible.
Chapter 1, though, sets the stage for our later studies, and as such does
not specifically map to any objective. For many of you, this first chapter may
be review. That’s excellent, and the authors won’t mind at all if you move
quickly through this information—or any topic where your current level of
experience makes you extremely confident. This chapter will help get you get
started on achieving your goal of the Server+ Certification.

What Is a Server?
W hen preparing for the Server+ exam, one question needs to be
gotten out of the way immediately: What exactly are servers, and what
makes them special enough to deserve an entire exam dedicated to them?
The answer to this requires that the term server itself be defined. Put simply,
there are two key definitions of server in the Information Technology world:
serv·er (sûrvr), n. 1. Computer software designed to assist other
computers on a network by performing tasks for them or providing
information to them. 2. Computer hardware optimized for the task
of running server software.
Each of these definitions needs to be considered separately, along with
its implications for what a “server” is. We’ll take some time in the following
sections to dissect these definitions, taking care to examine servers as
software as well as servers that operate solely as hardware. We will cover
scalability versus expandability, the relationship between security and

Copyright ©2002 SYBEX, Inc., Alameda, CA

What Is a Server? 5

dependability, and finally stability and redundancy. Before we move into

discussing the various roles of a server, we will overview server add-ons.

Server as Software
Let’s start by examining the first definition. For any computer to function, it
needs an operating system (OS). This is the code that tells the computer how
to function. You know that, of course. You have also probably encountered
the term NOS (network operating system), which is used to describe a server
OS. Things become a bit tricky, though, when we start trying to distinguish
an OS from a NOS.
The reason for this is that by the definitions we’ve just shown you, any OS
that can perform services or share files on the network is a server. Many
of you have used the file sharing capabilities of Windows 98, for instance.
All of Microsoft’s modern OSs have the ability to share out files, and even
to maintain NetBIOS browsing lists that allow computers to find each other
on the network. Even so, we don’t generally think of Windows 98 as a
“server OS,” and neither does the Server+ exam. Rather, the NOS term is
reserved for products such as Novell NetWare, Microsoft Windows 2000
Server, or Sun Solaris.
In order to decide which software you will need as your NOS software,
you will need to examine and consider the following characteristics:
 Client prioritization
Reviewing each of these characteristics in full is a good starting place
when considering server hardware for your NOS. As such, we will start by
examining the concept of scalability and how it relates to server performance.

Most computers serve only a single master, in that the user working locally on
the machine is the only one giving orders. The user may run one application,
or a number of them, but the amount of computer power a single user needs
is relatively limited, especially as we enter the world of multi-gigahertz pro-
cessing on the desktop. Because only one user is expected to be using the OS
at a time, a normal OS is intended for use on machines with limited resources.
Windows 98, for instance, cannot recognize or use more than one processor.

Copyright ©2002 SYBEX, Inc., Alameda, CA

6 Chapter 1  Server Types and Roles

Windows 2000 Advanced Server, on the other hand, supports 8 processors,

and Sun’s Solaris supports up to 128 processors on a single system.
Besides just allowing for more hardware, network operating systems are
designed to allow for features like clustering and load balancing:
Clustering Used to allow a number of servers to share resources,
clustering essentially creates a single “virtual server” out of a number of
machines. The computers share an IP address and generally use the same
data array, as shown in Figure 1.1.

FIGURE 1.1 Server clustering


Server Server


Load Balancing Similar to clustering, in that two or more servers team

up to do a single job. The thing that distinguishes load balancing, though,
is that each server retains its own identity and often keeps its own copy of
needed resources, as shown in Figure 1.2.

FIGURE 1.2 Server load balancing


Server Server

Data Data

Copyright ©2002 SYBEX, Inc., Alameda, CA

What Is a Server? 7

Scalability, then, is the ability of a computer system to support large num-

bers of users, and run extremely demanding software applications. That is
just one job for a NOS, though. Next comes the ability to keep things safe.

Network operating systems also are generally far more secure (or at least
securable) than client operating systems. This enhanced security can take the
form of a username/password database, access restrictions on files or
services, or any number of system security policies.
One of the odd things about the Server+ Exam is that, because most
of the questions follow a generic format, and because very little security
information falls into this “generic” category, you will find few system
security–related elements on the exam. This is strange, of course, because
network security is among the primary job functions of a server administrator!

The physical security of the server, however, is a major concern of the exam.
Locking down the server room will be dealt with in Chapter 13, “Managing
and Securing the Server Environment.” Some general security topics will
also be considered on a NOS-by-NOS basis in Chapter 7, “Network Operating

While most desktop PCs are shut down each night, and are used only a few
hours each day, servers are generally on 24/7, and as such they need an
OS that is extremely stable. Moreover, as tens or hundreds of people are
interacting with the server each day, it is critical that the OS be resilient and
able to deal with this constant onslaught of requests without locking up or
giving up.
To help guard the health of these machines, NOS software is often pickier
about what software it allows to run, and which applications and drivers
it will allow you to install. While this helps to insulate the server from prob-
lems caused by bad software, it also means that NOS applications often are
specifically written for the OS, and can be extremely expensive.

Client Prioritization
One last characteristic of a server OS is that it gives priority to client
connections when allocating resources. The primary purpose of a NOS is to
take care of clients, and as such a user at the server console is treated as just
another user, or sometimes even given a lower priority than network users.

Copyright ©2002 SYBEX, Inc., Alameda, CA

8 Chapter 1  Server Types and Roles

Some examples of server-class operating systems are listed below, along

with website information so you can go to learn more about each of them.
 Windows 2000 Server:

Solaris is a Unix-based system, and Unix operating systems are all based on a
server-class platform. Linux is also Unix based. Still, both Linux and Solaris are
often used as a desktop OS, and are as flexible as Windows in that they can be
used for pretty much any role in the enterprise. Start at
to learn more about the emerging Linux challenge to the established NOS/OS

Servers as Hardware
The second definition of a server is one that involves specialized hardware
designed to handle the extreme demands of NOS software and network
users. Companies such as IBM and Compaq produce computers specifi-
cally for these needs, and sell them in separate product lines. Compaq, for
instance, has its extremely popular Proliant series, and Dell sells the Power-
Edge line. At a very general level, servers are essentially just enhanced PCs.
Many managers look at the price of a new Compaq Proliant 1GHz server
and say, “Why are we paying $10,000 for this computer, when we could get
a Compaq PC that is just as fast for $1,000 at Circuit City?”
This is a valid question, because Windows 2000 Advanced Server or Sun’s
Solaris can be installed on a desktop-class PC without any trouble. If you are
in the position of proposing a server purchase to a manager or client, you
should be prepared to explain the reasons behind the higher cost of special-
ized server hardware.

It is worth emphasizing here that server-class software can be installed on a

desktop PC and that Windows 98 can be installed on an IBM server. To get
best performance, though, both the hardware and software need to be

Copyright ©2002 SYBEX, Inc., Alameda, CA

What Is a Server? 9

To help you with that explanation, let’s take a closer look at the benefits
a server provides for that extra money:

One of the most important characteristics of server-class hardware is that
it is generally built with generous expansion capability. Most servers allow
for far more RAM (often over 4GB), more drive space (most servers have
5–10 drive bays) and more processors—it is hard to get a desktop PC that fits
8 processors because the cases for normal PCs simply do not have room for
that much hardware. Along with all of this additional hardware comes
the need for additional fans and a larger power supply as well, which also
take up room.

Server hardware needs to be reliable. Unlike desktop PCs, which are gener-
ally shut down each evening, servers often are expected to run constantly for
weeks or months. The length of time a server has been running, or sometimes
the percentage of time it has been running, is referred to as its uptime. Some
servers prominently display the amount of time they have been up on their
console, while others (Windows, anyone?) tend to hide that information!
Any time a server is not running, the dreaded word downtime is used
to describe the amount of time that it is off. Because servers are critical to
modern networks, and networks are critical to modern organizations, a
server down situation rarely goes unnoticed. E-mail doesn’t work, or users
can’t get to files, or “the Internet doesn’t work,” and calls start flooding into
the help desk.
Along with backup and security, the prevention of downtime is probably
one of the most important jobs of an administrator. Server-class hardware
helps to maximize uptime through higher quality hardware and the ability
to duplicate critical hardware for redundancy.

One of the reasons servers cost more than desktop PCs is that the pieces used
to build the server are better. No one argues about why a Porsche costs more
than a Yugo, but somehow a lot of people who drive very nice cars find it
difficult to understand why they should pay for quality in their server room
as well.
Server components are manufactured to higher standards, both in terms
of the materials used and the precision of the craftsmanship. Moreover, these
components are tested to ensure that they work well together. This is done

Copyright ©2002 SYBEX, Inc., Alameda, CA

10 Chapter 1  Server Types and Roles

to ensure that a server will remain operating and reliable regardless of the
amount of work required of it. Much of this information is very different
from the way those same resources are discussed in the A+ book, which deals
with the maintenance of “normal” desktop PCs.

Quality components are great, but even the best machines sometimes
fail, and computers are no exception. In order to try to prevent hardware
problems from resulting in immediate downtime, though, most server-class
computers support redundant hardware for key components. This practice
is known as redundancy.
Redundant components can include power supplies, for instance. If a
server has two or more power supplies, both of them can work together to
power the system. However, if one of them fails (or is unplugged), the other
is able to take on an increased load and power the entire system. Other
examples of commonly duplicated hardware include hard drives, drive
controllers, and network cards.

Two items that are not redundant are processors and RAM modules. Even if
you have four processors in a machine, if one processor fails, the server will
go down. The same with RAM. Remember that expandability and redundancy
are different things!

Server-Only Features
Besides just supporting more and better hardware, and offering helpful ser-
vices not available on regular operating systems, modern servers also can be
equipped with a dizzying array of add-on equipment. Although some of these
components are making their way into the desktop computer environment,
they are normally associated with server environments. These include RAID
controllers (standard on most servers), SCSI controllers, an uninterruptible
power supply (UPS), external drive arrays, fax or modem bank hardware, and
tape backup drives. Any of these can be installed into desktop-class machines
as well, but generally their expense and resource requirements dictate that
they be used in server-class machines with server-class OSs.
You may already be in charge of a network, and have your own server(s)
to refer to as you read this book, but if not it might be useful to get an idea
of what these machines are like. Before reading too much further, you may
want to take a look at some of the beasts that the Server+ certification

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 11

prepares you to deal with. Below is a sampling of some large computer

hardware manufacturers, and their server lines:

While this book is being written, HP and Compaq are in the process of
merging, so things may be changing a bit there. For now, though, they
have separate product lines.

Server Roles
Servers must perform a dizzying variety of tasks on the network.
On smaller networks a single machine might perform many or all of these
tasks, and that is perfectly workable because servers are designed to be
good at doing multiple tasks simultaneously. On larger networks, though,
specialization allows each machine to be tailored specifically for the tasks
it is assigned.
This section will sample a few of these tasks for you and give you an
overview of what servers do on a network and how each of these tasks takes
its toll on server resources. Three general types of server roles will be detailed
in this section: security, network, and user. These are loosely grouped, and
some server roles cross over between the categories, so concentrate more on
what they do than where they are grouped. As you read through this section,
keep two questions in mind:
1. What roles does this server play on the network, and how does its
performance impact network users?
2. What type of operating system and hardware should be used to
improve the efficiency of the machine running this task?

Copyright ©2002 SYBEX, Inc., Alameda, CA

12 Chapter 1  Server Types and Roles

Security Roles
Networks have evolved into the storage location for almost all documents
and data in most large and midsize companies. The protection of this data—
both from destruction and from unauthorized access—falls to the server
Network security is generally provided by the operating system, and as
we look through each of the following sections, you will be given pointers to
websites that detail how these services are implemented into various server
systems. Because the Server+ exam itself is vendor-neutral, you don’t have
to spend a lot of time studying this stuff, but if you are interested in learn-
ing more about a topic, some of these URLs could come in handy as a
starting point.

Account Management
Most of the following services depend on the ability of the server to deter-
mine one fact—who it is that is trying to use the service. This is generally
accomplished by using one or more servers on the network to store and
authenticate user credentials.
Account management servers generally keep track of (among other
things) two basic pieces of information—the user and the password.

User (or Username)

The user is the basic building block of network security. The user defines an
individual (for example, “John Doe” becomes “JDoe”) or a network role
(“administrator” or “root”), and comes in one of three basic types:
Administrative At least one top-level account must be created during the
server installation process. That account is given authority to manage
the server. Sometimes called the “god account,” this first account is
known by many names by the various computer tribes. Microsoft calls it
the Administrator, Novell calls it Admin, and Unix/Linux calls it Root.
User Created The administrator can then create individual system
identities in order to differentiate between the various individuals on
your network, all of whom have their own network resource needs and
security clearances. Creating a user account registers the user as a security
object on the server.
Guest/Anonymous Sometimes access to the server needs to be granted
to users who are not registered in the account database. This can be done
by creating a default user. Anonymous users are most commonly seen on

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 13

Internet web servers, where thousands of users might visit a server to view
web content. The users do not have individual accounts, and they all log
on through a single generic account. FTP servers also often allow file
retrieval access to anonymous users.

It is almost inconceivable that anyone reading this book needs to have pass-
words explained. If you do, all I can say is you have a bit of work ahead of
you! If a username says “this is who I am,” a password says “and here’s proof
I am who I say I am.” Password security is a critical part of server security,
because if your account passwords—especially administrative passwords —
are discovered, the server and network security are compromised.
Many of the services listed in the authentication section that follows are
specifically designed to protect system passwords. Sadly, even the best secu-
rity software cannot defend passwords scribbled on sticky notes and posted
on a user’s monitor. An effective password strategy needs to include a pass-
word scheme that balances security and usability.

Windows server administrators can use a nifty tool called L0phtCrack

(recently renamed to LC3) to test their system’s defenses. LC3 and other
and eye-opening toys for various OSs are available from @stake at

Designing a Password Strategy

You are in charge of a committee studying the current password structure

on your company’s network. The purpose of the committee is to examine
the corporate security needs and see whether the existing password struc-
ture is sufficient.

To your dismay, the committee reports back that not only are password
standards not uniform throughout the company, but that many offices
do not have any password policy at all. In attempting to craft a password
strategy, you find the following:

 All users have Internet access.

 Many users are logging on and working from home.

Copyright ©2002 SYBEX, Inc., Alameda, CA

14 Chapter 1  Server Types and Roles

 The corporate intranet and FTP servers both support plain text

 Users often share machines, and many times they use a shared network

Obviously, this is a worst-case scenario. Still, the solution is relatively sim-

ple. Because you want to secure the network without causing undue trouble
to users, you could recommend that some of the following changes be

 Require a minimum password length.

 Require that passwords be changed monthly.

 Do not let users reuse old passwords.

 Change website and FTP settings to require encrypted logons.

 Require that all users have their own account.

This example is representative of the types of questions you might get. All
of the problems and solutions are generic because getting specific would
require mentioning particular authentication methods or require you to per-
form tasks on a particular directory structure. This would require dealing
with a certain vendor, and in most cases the exam is more interested in giving
you logic puzzles than it is testing your knowledge of a particular product.

The process of submitting a username/password set and having it tested
against credentials stored in a server database is called authentication. There
are a number of methods of authentication available to a server. Here is a
sampling, arranged roughly in order of least secure to most secure:
Plain Text This is the simplest form of authentication. In plain text
authentication, username and password information is simply sent out
over the network in clear text—standard ASCII code that can be inter-
cepted and read easily. Plain text authentication is highly frowned upon
in secure environments. Scratch that...plain text is highly frowned upon
for any environment.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 15

Encryption The process of protecting authentication information gener-

ally involves encrypting the username and password information as it is
transferred between the client and server. Encryption takes many forms,
but Solaris, for instance, supports Kerberos, Diffie-Hellman, and others.
Go to for
more on this topic.
Smart Cards Security can also have a physical component, and smart
card logon options are becoming more common. These depend on the
user knowing a piece of information (password) and having physical
proof of identity (generally a swipe card).
Biometric Identification The most complex and futuristic authentication
method—biometrics—uses retinal scans, voice scans, fingerprint analysis,
and other tools to provide physical proof of identity. The key is using a
human feature that is unique to each individual. Biometric ID integrates
physical identification with authentication.

The Benefits of Security

Many people take security for granted. After all, what’s the big deal? “There
is nothing on my computer worth stealing.” These are the famous last
words of many who have been caught up in an incident where their com-
puter has been broken into.

I recently experienced a situation that drives this point home. A private

school that I worked at long ago called me for some advice with regard
to their new network setup. The school had purchased a dedicated Internet
connection that was being shared to an entire lab of computers. Unfortu-
nately the server was not using a network operating system. The faculty
elected instead to use a desktop operating system. Desktop operating
systems do not provide the necessary level of security or authentication
needed for a server role.

In advising them on their network questions, I also cautioned them on the

security holes that their Internet connection and operating systems were
experiencing. Unfortunately it was only a few weeks later that I received
another call from the school: This time, they were frantic. They had received
a nasty letter from their Internet Service Provider, who was accusing them of
illegal activity. The faculty was completely oblivious to what was going on.

Copyright ©2002 SYBEX, Inc., Alameda, CA

16 Chapter 1  Server Types and Roles

After I assisted them with some research as well as checking into the log
files with the ISP, we discovered that a hacker had broken into their network
and was using their high-speed connection for illegal activities. Since the
operating system was not providing adequate security, this hacker easily
broke in and then masqueraded as a computer within the network. Never
take your network security for granted.

Directory Services Server

Directory services servers allow the processes of account management
and authentication to be handled centrally by a single machine or group
of machines. This allows a single username and password set to be used
throughout an entire network. Microsoft made the idea of “one user, one
account” into a mantra when they were promoting their Windows NT
domain structure, and the key premise is sound. Wherever a user goes in the
enterprise, that user should be able to authenticate to the network using
the same username and password. Figure 1.3 shows how a user can log on
in two separate areas of a Windows 2000 domain using the Windows 2000
Active Directory.

FIGURE 1.3 User logon to the Windows 2000 Active Directory

If credentials are
successfully checked,
Credentials checked the server provides
by the AD server resources
User enters

Windows Active Directory SQL Server

Network Client Server Database server

Both the Active Directory Server and the SQL Server run copies of the
Active Directory and share the user database as well as the responsibility for
authentication. Common directory services servers include the following:
 Microsoft’s Active Directory
 Novell’s NetWare Directory Service (NDS)
 Sun’s Solstice

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 17

The three services listed here are all based on the standards set by the Inter-
national Telecommunication Union ( The ITU’s x.500 directory
specification defines how accounts are created and managed. Because of
this, all three of these systems use similar structures and logic, making it
far simpler to manage multiple systems than ever before.

Security servers do not need to be tremendously powerful in terms

of hardware, but they must be powerful enough to respond quickly to
client requests. If the directory server is unable to keep up with authenti-
cation requests, the network simply slows down to its speed. Nobody gets
anywhere without permissions, and these servers are the gatekeepers.
If we were to pick just one resource to emphasize, it would probably
be network throughput, which is critical to a security server. Being able
to receive and send requests quickly can be facilitated by multiple network
cards, or even by locating the directory server in a central part of the net-
work. Also, encryption technologies can be heavily processor intensive, and
so servers with enhanced authentication schemes may require additional
processor power.

Networking Roles
Security work is the glamour job in the server world, and Active Directory
and Kerberos (a network security system, developed at MIT, which
verifies that a user is legitimate at login) seem to get all the attention
in the trade rags. Still, in order to make a network function smoothly,
a number of other services also need to be working in the background.
These services assist the network in locating servers, identifying computers,
connecting remote clients, or moving packets from one part of the network
to another.
You won’t be asked to know about the specifics of any one of these
technologies, since each of them is implemented a bit differently by different
server platforms, but you should be familiar with what they do, and the
basics of how they work.

Copyright ©2002 SYBEX, Inc., Alameda, CA

18 Chapter 1  Server Types and Roles

Routing Services
One of the features that a server can offer to the network is to act as a router.
Routing and bridging services allow a server with multiple network interfaces
to link machines on either side. When acting as a router, the system must
build a routing table that shows which machines are available on which
As TCP/IP is by far the most common protocol you will need to deal with
(some NetWare environments still use IPX/SPX), the IP routing protocols are
the most important ones to keep in mind:
RIP The Routing Information Protocol (RIP) is a distance-vector pro-
tocol that enables computers to exchange routing information by means
of periodic routing table updates. RIP updates are sent to neighboring net-
works and RIP information from other routers is returned. The path with
the fewest hops (each router involved in a path is one hop) is used when
sending data.
OSPF Open Shortest Path First Protocol, or OSPF, is an open protocol,
meaning that it is a standard and that it is available for use in the public
domain. OSPF is a link-state routing protocol. Link-state advertisements
(LSAs) are sent to all other routers to allow them to update their routing
tables. These LSAs include changes to the routing table, but the actual
routing table itself is not sent, unlike RIP. OSPF is far more efficient than
RIP on large networks.

Although server machines can be used as routers, in most cases it is better to

purchase a dedicated router for this job. Cisco ( and 3Com
( are two major router manufacturers.

Firewall Server
A firewall is essentially a router turned bouncer. Firewalls are placed at the
edge of your network and are used to turn away communications from
unwanted or distrusted clients. Nearly every large corporate or organiza-
tional network now has at least one high-speed connection to the Internet.
While this makes it extremely easy for network clients to access the Web
and e-mail, the process also works in reverse, and networks are vulnerable
to attack from the Web. As such, nearly all firewalls are concerned with the

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 19

need to protect a local area network (LAN) from the perils of the Internet.
Figure 1.4 shows a common network configuration and introduces a concept
we need to define.

FIGURE 1.4 A common firewall configuration

Firewall Server
Public Network

Gateway to Internet


Hub or Switch
Private LAN
Web Server


Note the DMZ, or demilitarized zone. The DMZ is the buffer zone
between the Internet and your internal network. It is where any servers that
need to be exposed to the Web should be housed. In this case, a web server
is sitting in the DMZ, and a server running firewall software protects the
intranet. Any requests sent to the web server—including malicious DoS
(denial of service) attacks or Internet worms—will be able to reach that
server unhindered. The same requests or attacks directed toward the internal
network, though, will be intercepted by the firewall, which will be config-
ured to allow only particular information through.

In many cases, it is best to put the web server behind the firewall as well. The
firewall can let through HTTP requests while protecting the server.

Proxy Server
A special kind of Internet access server is a proxy server, which is part
accountant and part traffic cop. Proxy servers are used to funnel all Internet
traffic through a single location, and because of this central point, they can
effectively manage Internet traffic. Notice in Figure 1.5 that clients on the

Copyright ©2002 SYBEX, Inc., Alameda, CA

20 Chapter 1  Server Types and Roles

network direct all requests for Internet information to the proxy server
\\Trantor. The server then checks for a number of things:
1. Is the user allowed Internet access?

2. Is the user allowed to go to the intended website?

3. Is the web page already requested in cache?

FIGURE 1.5 A proxy server at work

Client 1 Client 2 Client 3

Client requests
a web page

Server checks to see if

the user is authorized
to view the page. If they Internet
are, it is retrieved either
from a local cache or
from the Internet. Trantor

Although a standard firewall can perform the first two of these tasks, the
proxy’s ability to cache pages for users makes it invaluable in saving on lim-
ited bandwidth. If three users request a page, only the first user’s request
actually hits the Internet—the other users are then given the page from inside
the cache.
One service that both firewalls and proxy servers provide is logging of
user requests. Administrators can parse the log files for any of the key unac-
ceptable words or phrases, or just check to see who is downloading .mp3s
at work.

Most of us didn’t sign up as network engineers with the intention of becoming

morality police or productivity enforcers. Still, it is an unfortunate fact that this
is a part of the job on modern networks.

Most routing services require little in the way of resources, with extremely
fast network connections being the key to their success. Making sure that
these machines have quality network cards and up-to-date network card
drivers is critical. Also, because the proxy server caches large amounts of
data, it may need a large drive array.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 21

For a great look at how routers, firewalls, and proxy servers work, check
out the excellent film, Warriors of the Net. It is available to view free at, but you will need a fast connection, as the
high-res version is 150MB. It’s worth the time and the wait, though.

Remote Access Services

Ten years ago, a user leaving for a business trip must have felt an amazing
sense of freedom. No cell phones! No laptop! Check in by phone every day
or so, and otherwise you were safe from whatever disasters might need your
attention at the office.
Not so anymore.
As computer-based tasks become a more critical part of the normal work-
day, networks have evolved to provide easier access to resources for users
who can’t get to their desks. Among the most important tools available to
us who use laptops are remote access servers. These are servers that allow
remote users to function on the network as though they were in the office—
even if they are thousands of miles away!
Two of these are the traditional dial-in server and the virtual private net-
working (VPN) server. Each of these has advantages and drawbacks.

Dial-In Server
A dial-in server is essentially a router that has a modem as one of its network
interfaces. The server answers calls coming in from remote clients, authen-
ticates them with the network, and then acts as a conduit, allowing them to
access resources on the network.
In order for this process to work, a dial-in protocol must also be available.
The most common of these are Point-to-Point Protocol (PPP) and the less-
used SLIP (Serial Line Internet Protocol). PPP is newer, and is generally more
efficient because it has error-checking mechanisms built into it. For more
information on how PPP and SLIP differ, check out
survival-kit/slip-vs-ppp.html for one ISP’s explanation of the two.
SLIP is not used much anymore, but it is good to at least recognize the acro-
nym, just in case.
The hardware requirements of a dial-in server can be quite specialized
because the task of providing many—often dozens—of modem connections
is beyond the comm port capabilities of a standard machine. Specialized
expansion boards by companies like Digi ( allow a server to
support this higher hardware level.

Copyright ©2002 SYBEX, Inc., Alameda, CA

22 Chapter 1  Server Types and Roles

In Figure 1.6, a Windows 2000 Server is using the Routing and Remote
Access Service (RAS) to support dial-in clients. The clients authenticate to a
Windows domain controller on the network, and are then able to connect for
e-mail and file access.

FIGURE 1.6 A dial-in configuration for Windows

Server answers and

verifies the user's
identity and rights
with a directory server
Directory Server
Client dials
into RAS server

RAS Server
requests resources
Remote Server transfers RAS Server for the client
Client the data out to
the remote client
File server returns
the data to the server

File Server

VPN Server
A virtual private network (VPN) is similar to a dial-in connection in that it
allows users to access their network remotely. Unlike standard dial-in,
though, a VPN connection involves a two-step process:
1. Users attach to the public Internet using a dial-up or by configuring
their machine to use a high-speed connection.
Once an Internet link is established, users can start a VPN client to
make a connection across the Internet to a VPN server on their own
network. This server also needs to have a separate Internet connec-
tion. This process involves creating a secure tunnel connection through
the existing connection established in step 1. This secure connection is
called a Point-to-Point Tunnel Protocol (PPTP) connection.
The VPN connection is encrypted, and because all communication is
encapsulated within the VPN protocol, users can access network resources
through the VPN that they would otherwise be unable to see using standard
TCP/IP connectivity. Figure 1.7 shows a common VPN configuration. Note
how similar this process is to the one shown in Figure 1.6.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 23

FIGURE 1.7 Connecting through a VPN

Server answers and

verifies the user's
identity and rights
with a directory server
Directory Server
Client dials
into VPN server

VPN Server
requests resources
Remote Server transfers RAS Server for the client
Client the data out to
the remote client.
File server returns
the data to the server

File Server

TCP/IP Services
TCP/IP has been mentioned briefly already this chapter and as you go
through this book you will continue to hear about it. As earlier mentioned,
this is essentially the only major protocol standard that you can depend on
any server to support. Because it is everywhere, certain server functions
needed by TCP/IP networking must be included on nearly all networks.
We will just mention these here, as Chapter 8, “TCP/IP,” deals in-depth
with understanding and configuring TCP/IP networking.

DHCP Server
The Dynamic Host Configuration Protocol is used to simplify TCP/IP con-
figuration on network clients. DHCP servers store the information needed to
bring a TCP/IP client onto the network, and when a client first starts up they
contact the server to obtain an address, gateway, subnet mask, and DNS
server, among other things.

If these terms are not already familiar to you, Chapter 7 alone may not be
enough! In that case, Andrew Blank’s TCP/IP JumpStart (Sybex, 2000) is a
good reference. Vendor-specific TCP/IP books are also available, but the
JumpStart book is nice because it maintains the vendor neutrality that
CompTIA espouses.

The DHCP server itself can run on any platform, and clients from
multiple operating systems can use a single DHCP server. It is important to

Copyright ©2002 SYBEX, Inc., Alameda, CA

24 Chapter 1  Server Types and Roles

remember, though, that DHCP requests are done through network broad-
casts, so you generally need either a DHCP server or a DHCP relay agent on
each network segment. More on relay agents in Chapter8!

Name Server
Computers talk to each other quite happily using numbers, as digital infor-
mation all boils down to ones and zeros eventually. Human beings, on the
other hand, generally have an easier time with information presented to
them in the form of words and text characters. Because of this, name servers
allow both people and machines to have their own way. There are two pri-
mary types of name servers to keep in mind:
DNS Server The Domain Naming System (DNS) has been in use for
nearly two decades now, and is the worldwide standard for identifying
computers on TCP/IP networks. DNS servers resolve TCP/IP host
names to IP addresses. In Figure 1.8, the host Client1 requests access to The name is resolved by the DNS server, and Client1
can start the connection.
WINS Server Figure 1.8 also demonstrates the functionality of a WINS
server. WINS stands for Windows Internet Naming Service, and the “Win-
dows” part of that is a pretty good clue that this is not a vendor-neutral
service. Microsoft has been using its own naming structure—NetBIOS
naming—since the days of DOS, and a WINS server is used to support this
in a TCP/IP environment. Client1 can also access Server2, but does so
through a WINS server rather than a DNS server.

FIGURE 1.8 Name servers in action

Client 1

DNS Server WINS Server

Server 2

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 25

A configuration like the one in Figure 1.8, where both DNS and WINS servers
are in use, is common. These services can coexist, and even can help each
other out on occasion (see Chapter 7 for more on that).

Management Server
TCP/IP also provides a protocol specifically designed for network manage-
ment functions. The aptly named Simple Network Management Protocol
(SNMP) is used to allow a server on the network to collect information
about other devices and issue commands in return. We won’t spend a lot of
time on SNMP, as it is doubtful you will find detailed SNMP questions on
the exam. Even so, is a good place to go for an
overview of what SNMP is about.
In most cases, naming and management services are almost unnoticeable
in terms of their effect on the server. Even a very small server can support
thousands of clients with no problems. Again, the key is having sufficient
bandwidth to the server.

User Services
Finally we arrive at the services that users can see and interact with. When
most people talk about a server, they are concerned with the tasks discussed
below. This does not, of course, mean that they are the most important ser-
vices. Like a quarterback on a football team, user services get all the press
and most of the resources. The underlying network services listed above,
though, are as important and underappreciated as offensive linemen!
Notice that when services are started, the result is significant resource
usage. You should be able to identify how to plan for each of the following
types of servers by planning to boost critical resource needs.

File Server
The classic task of a network server is to store information that needs to be
shared among multiple users; in this role it is known as a file server. To suc-
cessfully store and share information, the server normally has to have a few
different elements in place. First, some sort of security needs to be present to
protect the files on the server. Different network operating systems handle
security in very different ways, but in all cases the server needs to be able to
ensure that users do not have access to files they should not see. Servers can

Copyright ©2002 SYBEX, Inc., Alameda, CA

26 Chapter 1  Server Types and Roles

also make more subtle distinctions, such as allowing users to read a file but
not modify it.
The requirements for a file server are heavily weighted toward its hard
drives. Server hardware often comes with multiple drives, because file servers
are expected to store enormous amounts of data. File servers also need fast
drives—Compaq uses 10,000RPM drives in its servers. The drive controller
is also important, as is the network bandwidth available to the server. In
Chapter 4, “Storage Devices,” we will examine server drive configuration,
and you will notice that SCSI hardware is the overwhelming choice for
servers. This is because SCSI is faster and more expandable than IDE/EIDE.

It is interesting to consider that a traditional web server is actually nothing

more than an Internet-based file server. Same with FTP servers. Web servers
receive requests for HTML pages and serve those pages out, just as a file
server sends out .doc and .txt files.

Print Server
If you were interning to be a server on a network, it is likely that you would
start as a print server. Print servers require very little in the way of resources,
outside of requiring sufficient drive space to store files submitted for print-
ing. Even this is a relatively small requirement, because print jobs are gen-
erally stored only until they are printed, at which time they are deleted. The
process of network printing is enumerated below. Two terms you should be
familiar with when discussing printing are queue and spool:
Queue A queue is a list of documents waiting to be printed. The term
also describes the location where these documents are held.
Spool Spooling is the process of writing a document into the queue. The
queue is often called the spool file in fact. Spooling allows a print job to
be sent to the server even if the printer is busy, thereby freeing up the client
to continue on other tasks.
1. The client chooses to print a document. Part of this process involves
choosing a printer (or accepting the default printer).
2. The client’s printer driver is used to format the document for printing
on the particular printer chosen.
3. The document is submitted as a job to a local print queue on the client.
This is optional, but as it immediately frees up the client PC to work
on other tasks, local spooling is pretty standard.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 27

4. The job is then sent from the local print spool to the network print
server. This server has another print queue, and the document is
placed here.
5. The print job is placed in line with the jobs of other users, and when
the printer is prepared to print it, the job is spooled out to the printer.
6. The printer produces the document, and reports back to the server,
which deletes the job from its queue.

It is possible to tell the print server to keep all print jobs rather than delete
them. While this is good for seeing what has printed, it can eat up drive space
and is not normally recommended.

A print server’s primary task is to interface with machines that are pain-
fully slow by computer standards—even the fastest printers move at a glacial
pace compared to PC speeds. Because of this, print servers require minimal
hardware, and print services can often be combined with other tasks rather
than having a dedicated print server.

Application Server
File and print servers are in many ways the backbone of a network—
application servers are its brain. App servers are machines running server
processes that perform tasks on behalf of users, or interact with client
machines in the completion of tasks.
There are a number of different application servers, but three of the most
common are these:
 Database server
 E-mail server
 Active web server
The key to a server being classified as an application server rather than a
file server has to do with how much work the server does on the data before
sending it to the client. A great example of this can be found in Microsoft’s
database family.
Microsoft Access is a database program that can share a database among
multiple users. Because of this, the Access data file itself can be placed on a
server and made available to network users. At that point, the server is shar-
ing out a database, but it is not an application server. The reason for this is

Copyright ©2002 SYBEX, Inc., Alameda, CA

28 Chapter 1  Server Types and Roles

that if a client requests information from the database, the process shown in
Figure 1.9 is initiated.

FIGURE 1.9 Requesting data from an Access database


Access Client Client Requests Data

File server returns entire database


Notice that the client needs only a specific set of data, yet the server sends
the entire database across the network to the client, which is then responsible
for sorting out what it wants and discarding the rest of the information. This
is inefficient in two critical ways:
1. Time and bandwidth are wasted transferring unneeded rows of data.

2. Processing of the query is done on the less powerful client, meaning it

will take longer to complete the task.

Client-Server Architecture
The solution to this problem is the use of a client-server architecture, such
as the one available in Microsoft’s SQL Server. Client-server applications are
computer programs that are specifically designed to use the processing
power of both the server and the client machines in the completion of their
tasks. Generally this means that the client makes an initial request to the
server, and the server then does some initial processing on the request. The
result of that processing is then returned to the client, or to another machine
for additional work to be done with it.

If more than just a single client and server are involved, this is called an
“n-tier” architecture; n stands for the number of machines used in the
processing, meaning you could have a client and two servers involved
in the transaction, and it would be a “3-tier” design.

Microsoft SQL Server is a server-side application that runs as a service

and works with clients to ensure that they are given only the data they

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Roles 29

Figure 1.10 shows the same request being issued by the client as in
Figure 1.9, but with a significantly different response.

FIGURE 1.10 Requesting data from SQL Server

Client Requests Data Server

Access Client
Server accesses
database and
processes query

File server returns a requested subset

of the database

Do you see how this time the server has actually looked at what the client
needs and has preselected the data? By doing this, both the network and
the client are less heavily taxed, and the server is able to justify its expensive
hardware by actually doing something.
Because a large database server may be doing tasks for dozens—or even
hundreds—of users all at once, the hardware requirements on an application
server can be extreme. Moreover, because app servers do a lot of “thinking,”
faster processors or multiple processors can be crucial.

The Carnegie Mellon Software Engineering Institute has a nice reference

on client-server technologies at

Internet Server
The last server type we will consider is one intended to deal specifically with
web-related or other Internet-related client requests. A number of Internet
services can be provided by network servers, but probably the most common
of these are the web server, the mail server, and the FTP server. Increasingly,
though, streaming media servers, online database servers, and Internet-
specific application servers are coming into use.

Web Server
The World Wide Web started out as a collection of HTML (HyperText
Markup Language) pages stored on Internet servers. Over the past few years,

Copyright ©2002 SYBEX, Inc., Alameda, CA

30 Chapter 1  Server Types and Roles

though, web servers have gotten progressively more complex, and HTML
has evolved from a static file server technology to an active client-server
The interaction between web servers and web clients (browsers) has
now become quite complex. Java, ActiveX, server-side scripting, and
database connectivity through the Web have all increased the power
and potential of web servers. Many enterprises now find that their web
servers are an integral part of both daily business environment through
intranets and web-based applications.

FTP Server
FTP servers, on the other hand, remain very much the same today as
they were 10 years ago. An FTP server is just a file server for the Internet,
operating over the FTP protocol. Clients connect to the server, authenticate,
and add (PUT) or retrieve (GET) files just as you would on any file server.
The hardware requirements for web servers are fluid, as these servers can
support a few concurrent (simultaneous) connections or a few thousand. As
your expectation of the number of people using the site rises, so should your
hardware levels.

Mail Server
There are a number of different e-mail server options available for use with
your network. Most NOS vendors have e-mail packages available for
their server operating systems, and a number of freeware or shareware
e-mail servers are in use as well.
Besides providing for the critical ability to send and receive messages,
e-mail servers can filter out inappropriate messages, provide protection
from e-mail borne viruses attempting to enter the system, and act as a
repository of information and communication data for the organization.

Because they do so much more than just shuffle mail around, these applica-
tions are often referred to as groupware rather than just as e-mail servers.

In this chapter, we have discussed what servers are and how to identify
server-class hardware and software. Knowing how to tell what hardware
components are appropriate, and which operating systems are designed for

Copyright ©2002 SYBEX, Inc., Alameda, CA

Exam Essentials 31

server work, is critical when you are choosing a new server or deciding
whether an existing box is up to a new task.
We also looked at a sampling of the jobs that servers do, and examined
what types of hardware are needed for certain tasks. If you haven’t already,
spend a bit of time browsing the Internet links associated with the topics in
this chapter. There is a lot of good information there, and Web data hunting
is among the most important skills you will need to develop as a server
Throughout the rest of the book, you will take a Chapter Review Test.
Each test will consist of 20 questions designed to quiz you on the objectives
and content that you reviewed within the chapter. As stated, this chapter
does not cover any particular exam objectives—but, to keep your test-taking
skills sharp, we included 20 questions to reinforce some of the material
you’ve just reviewed. Much like the Assessment Test you took in the Intro-
duction, this test will help you target areas you may need to refresh before
forging ahead with the exam preparation. Good luck!

Exam Essentials
Know what a server is. Servers can be hardware or software that
provides a service for other devices connected to the network.
Know the characteristics of server operating systems. This includes
scalability, security, stability, and client prioritization.
Know the benefits of using server hardware. Expandability, depend-
ability, quality, and redundancy are the benefits of using server hardware
over server software.
Be familiar with common server roles. Servers can perform the
following roles within a network: security (account management,
authentication) and directory services.
Know the major routing protocols. RIP and OSPF are the main routing
protocols used today.
Know what a proxy server is. Proxy servers perform Internet tasks on
behalf of the computers on the network.
Know the main types of remote access servers. This includes dial-in
servers and VPN servers.

Copyright ©2002 SYBEX, Inc., Alameda, CA

32 Chapter 1  Server Types and Roles

Be familiar with the different types of user services that a server can
perform. Servers can fulfill the following user services: file server, print
server, application server, Internet server, web server, FTP server, and
e-mail server.

Key Terms
Before you take the exam, be certain you are familiar with the follow-
ing terms:

application servers operating system (OS)

authentication Point-to-Point Protocol (PPP)
biometrics proxy server
client-server queue
clustering redundancy
directory services Routing Information
servers Protocol (RIP)
downtime server
file server smart card
firewall spooling
load balancing spool file
network operating system (NOS) uptime
Open Shortest Path First virtual private network
Protocol (OSPF) (VPN)

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 33

Review Questions
1. Which technology allows a number of servers to share resources and
create a single virtual server out of a number of machines?
A. Failover

B. Clustering
C. Scalability

D. Mirroring

2. Which server is used to funnel local Internet requests through a

single location?
A. Proxy server
B. Firewall server

C. VPN server

D. Directory Services server

3. Which of the following is not a reason for purchasing server hardware:

A. It is less expensive.

B. Multiprocessor support.

C. Expanded software support.

D. Expanded hardware support.

4. What is a key problem with using a desktop PC as a server?

A. Can’t install NOS software.

B. Unable to add security features.

C. Case does not have space or power capacity needed.

D. Nothing. Desktops are the recommended server platform.

Copyright ©2002 SYBEX, Inc., Alameda, CA

34 Chapter 1  Server Types and Roles

5. Which term is used to describe a server failure that causes users to

be unable to use the system?
A. Restore

B. Backup
C. Uptime

D. Downtime

6. Which type of password authentication is the least secure?

A. Kerberos

B. Plain text

C. Smart cards

D. All are equally secure

7. x.500 is a standard of ___________.

A. Microsoft

B. CompTIA



8. Which of the following are routing protocols?




9. PPTP is used with what type of network service?

A. DNS resolution
B. Authentication

C. Virtual private networking

D. Firewall access

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 35

10. A DNS server is what type of server?

A. Application

B. File

C. Naming
D. Remote Access

11. What is a server (select all that apply)?

A. Computer software designed to assist other computers on

a network
B. Computer hardware optimized for the task of running server
C. A computer within a network that is used to perform advanced
network calculations
D. A user’s computer

12. Which of the following is not a priority consideration in deciding on

a network operating system?
A. Scalability

B. Security

C. Ease of administration

D. Stability

13. What is clustering?

A. Grouping of client computers together on a network

B. Using more than one server on a network

C. Grouping servers together to share resources for users

D. Linking servers together to share work loads

Copyright ©2002 SYBEX, Inc., Alameda, CA

36 Chapter 1  Server Types and Roles

14. What is load balancing?

A. Using more than one server to perform a single job

B. Using more than one server on a network

C. Sharing a single server across multiple resources

D. Adding more than one component within a server to prevent single
component failure

15. What is considered the basic building block of network security?

A. Users

B. Servers

C. Client operating systems

D. Network operating systems

16. Which of the following is not a secure password recommendation?

A. Requiring a minimum password length.

B. Requiring passwords change monthly.

C. Not letting users use old passwords.

D. Maintaining the same passwords locally as through remote access.

17. Which of the following is not a form of authentication?

A. Biometrics

B. Encryption
C. Write-protect tabs

D. Plain text

18. What type of protocol is RIP?

A. Distance vector
B. Link state



Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 37

19. What is a queue?

A. A list of printing protocols

B. A type of printer driver

C. A list of documents waiting to print

D. A pathway between a computer and a printer

20. Which of the following is not a type of application server?

A. Database server
B. E-mail server

C. Web server

D. TCP/IP server

Copyright ©2002 SYBEX, Inc., Alameda, CA

38 Chapter 1  Server Types and Roles

Answers to Review Questions

1. B. Clustering allows two or more PCs to act as a single server,
providing higher availability and performance than a single PC could
handle. They can share an IP address and use the same data array.
2. A. A proxy server is a specialized Internet access server. Just as the
word proxy implies, it acts as an intermediary between a client and a
server. Because proxy servers are used to funnel all Internet traffic
through a single location, they can be extremely effective in managing
Internet traffic. This central point allows the firewall software to filter
requests more easily.
3. A. Server hardware is expandable and flexible, but it isn’t cheap. The
expense of server hardware is the key hangup with purchasing server-
class hardware in many organizations.
4. C. Although it is possible to use a standard desktop PC as a server,
these machines are generally not designed to handle the multiple drives,
multiple processors, or large amounts of RAM that servers need.
5. D. Uptime is how long the server has been running since its last shut-
down, while downtime is the amount of time the server is unavailable.
6. B. Plain text passwords are not encrypted, and can easily be inter-
cepted by others on the network. Kerberos and smart card technolo-
gies are both encrypted, and are far more secure.
7. C. The ITU created the x.500 directory structure that is the basis for
the NetWare Directory Service (NDS), the Microsoft Active Direc-
tory, and other network directory services.
8. A, C. RIP and OSPF are both used to help routers build routing
tables. A DMZ is a border area between a public and private network,
and SLIP is a dial-up protocol.
9. C. PPTP and L2TP are both used with VPN access. The other options
are not directly associated with PPTP.
10. C. DNS stands for Domain Naming System, and DNS servers are
using for maintaining and resolving TCP/IP host names.
11. A, B. Servers can be either hardware or software that assist other
computers on the network.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Review Questions 39

12. C. Although ease of administration may be a consideration, it is not

a priority in deciding on a network operating system.
13. C. Clustering is grouping servers together to share resources for users
with redundancy.
14. A. Clustering is using more than one server to perform a single task.
This prevents one server from becoming overrun with requests.
15. A. Due to the fact that they control their own passwords, users are
considered the basic building block of network security. If users do not
keep passwords secure, the entire network security is jeopardized.
16. D. Allowing users to maintain a common password for both internal
and remote connections compromises network security.
17. C. Write-protect tabs are not a form of authentication but rather a
form of data protection.
18. A. RIP is considered to be a distance vector routing protocol.

19. C. A queue is a location where documents are kept in order until they
can be printed.
20. D. TCP/IP is not a type of application server. It is a network protocol.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Installation



 1.1 Conduct pre-installation planning activities

 Plan the installation
 Verify the installation plan
 Verify hardware compatibility with operating system
 Verify power sources, space, UPS and network availability
 Verify that all correct components and cables have been

 1.2 Install hardware using ESD best practices (boards, drives,

processors, memory, internal cable, etc.)
 Mount the rack installation
 Cut and crimp network cabling
 Install UPS
 Verify SCSI ID configuration and termination
 Install external devices (e.g., keyboards, monitors,
subsystems, modem rack, etc.)
 Verify power-on via power-on sequence

Copyright ©2002 SYBEX, Inc., Alameda, CA

T he two CompTIA objectives listed above concern installation,
and deal with the tasks and planning that must be done before your server
is even plugged in. CompTIA estimates that about 17 percent of the exam will
come from this material. For the most part this chapter is OS-free, and will
concern you only with the process of planning the server install and acquir-
ing, unpacking, and setting up your hardware.
Over the course of this chapter, we will deal with each subobjective in
turn, along with a number of thoughts on how such topics could turn into
exam questions. The hard part about these topics is that there are actually
very few “right” answers on how to plan a network, and many of the ques-
tions in this content area depend on you to make logical distinctions as to
what is “best” or “better,” not what is “right” or “wrong.” This makes for
some incredibly tricky (and frustrating) test questions, as you might suspect.

The two subobjectives not covered in this chapter are “Verify hardware com-
patibility with operating system” and “Verify SCSI ID configuration and ter-
mination.” SCSI is covered in detail in Chapter 4, “Storage Devices,” so we
will leave the discussion of termination and SCSI IDs until then, and operating
systems and their individual quirks will be considered in Chapter 7, “Network
Operating Systems.”

Plan the Installation

T he very fact that you are reading this book shows that you have an
understanding of the key point of this first chapter: A bit of hard work now
can save a great deal of trouble and disappointment later. Just as pre-test
studying is critical if you are going to be successful for the Server+ exam,

Copyright ©2002 SYBEX, Inc., Alameda, CA

Plan the Installation 43

so pre-installation work is critical if you are going to be successful in the

server room.
When beginning any new project, the engineer in charge needs to carefully
manage a number of different elements:
 Defining the project goals
 Examining the current configuration
 Budgeting the project
 Setting a timeframe for completion
Simply put, this amounts to the following: What do we want to accom-
plish, what do we have already, what do we need to acquire, and how much
time and money do we get to do it?

Defining the Project Goal

The first of these is always going to be the most important. Unless you go
into the project with a clear idea of what it is you are looking to do, much
of your effort is likely to be squandered on tasks that prove useless later on.
This does not mean, though, that the project goal needs to be extremely
detailed. Rather, it just has to be extremely focused. For instance, a good
initial goal might be the following:
 Purchase and install a database server for the company’s new
SQL-based accounting application.
Compare this to a far more detailed plan, such as the one below:
 Purchase and install a database server for the company’s new
SQL-based accounting application, and upgrade the server to
facilitate the new software and client needs.
 Upgrade network security and install a VPN to provide for increased
security so users can access the SQL server across the Internet.
 Back up the entire server each night and verify backups on a weekly
There is nothing wrong with any of the ideas in this plan, but the problem
is that this is really three or four plans, not one. Of course you will have
to back up the server, but the process of researching and buying a tape
drive will probably depend in large part on what options you choose for a

Copyright ©2002 SYBEX, Inc., Alameda, CA

44 Chapter 2  Installation

server OS, and what database you purchase. Moreover, making a declar-
ative such as “back up the entire server each night” implies that other
options—such as differential or incremental backups—have been auto-
matically removed from consideration. Further discussions on backups
will be covered in Chapter 14, “Backup.” A well-constructed goal should
be one that encourages ideas, not blocks them.
As that first goal from above seems as good an example as any for dis-
cussing planning, we will use this project as our example throughout this
chapter. This will allow you to see what some of the issues that might come
up in an actual install might be.

Examining the Current Configuration

First off, the engineer in charge of this project will have to examine the cur-
rent environment at the company. There are two major reasons to do this:
1. To find out what is currently available that can be leveraged or
cannibalized for the new project
2. To determine if any conversions or backward compatibility are
It is likely that the company already had an accounting system, and as
such you will need to determine what that system is, and what it is running
on. If you can continue to use the current hardware, this will of course save
considerably in terms of your budget. Other considerations, such as where to
put new hardware, or how to install the software, may be influenced by the
current network configuration as well.
As you are documenting the current configuration, there is one more
critical question to keep in mind:
Considering that there is already a server in place doing the job (in this
case it is the accounting server), why is this upgrade being done, anyway?
The way to answer that question, of course, is to talk to people. Find out
what the problem is with the current system. If users say it is too slow, you
may want to concentrate on making sure that the hardware for the new
server is even stronger than you would normally plan. If the upgrade is being
done to gain a particular enhancement, be sure you research how the imple-
mentation of that enhancement is accomplished on the new system. The key
here is that upgrades always mean change, and users generally resist change.
By assuring that the new system will immediately show its best side to the
users, you make it more likely that they will accept it, rather than grumbling.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Plan the Installation 45

First Impressions Are Everything

True story: I was involved in upgrading an office of 150 people from ccMail
to Exchange/Outlook a couple of years back. We had trained the users on
how the new system would work, tested all the server and workstation
software for compatibility, and the entire rollout crew was confident that
the new system would go in without a hitch. People went home on Friday
night, we migrated them over the weekend, and when they came in Monday
morning, all their data had been transferred to Outlook and everything
worked great.

Unfortunately, that day was also the day that the “Melissa” e-mail virus
debuted, and it hit our location at about 10 A.M. The entire network was com-
pletely brought to its knees within minutes. It took days to get things back to
normal, and by then most users were begging for their ccMail back. They
didn’t necessarily know why the system had failed, and they didn’t really
care…they just remembered that this hadn’t ever happened with the previ-
ous system. One oversight, in other words, can make you and your work
look really bad, and it takes a long time to overcome a bad first impression.

Once you have talked to users about the system, you may want to go back
and modify your original project goal. Perhaps remote access to the server is
one of the critical areas driving the upgrade. In such a case, you may modify
the plan to read like this:
 Purchase and install a database server for the company’s new SQL-based
accounting application, and upgrade the company web server to allow it
to host software that provides a web interface to the new database.
Part of the business case for upgrading the accounting server, in other
words, was to provide web access to data. This will be a key area that those
judging the success or failure of the project will examine, so you will want to
make sure that the hardware and software on the web server are up to the task.

Budgeting the Project

Once you are confident of what it is you need to do, the next step is to
sort out what additional resources you will need to complete the project.
Examine the list of what you have and what you need, which was compiled

Copyright ©2002 SYBEX, Inc., Alameda, CA

46 Chapter 2  Installation

in the last step. The administrators in the example we are using have come
up with the list in Table 2.1 (costs are as of Fall 2001, are not necessarily the
best prices available, and are used for example only).

TABLE 2.1 Preparing a Hardware Budget for the Upgrade

Current Recommended Cost to

Item Configuration Configuration Upgrade

Database server Pentium II 450 Dual P4 1.2GHz $12,000

hardware 256MB RAM 1GB RAM
27GB storage 80GB storage

Database NOS Windows NT 4 Windows 2000 $500, plus $40

Server Server per user for CAL

Database server Access 97 SQL Database $2,000–$10,000

software Server Depends on
license type

Database Custom Access SQL-based $50,000, plus

accounting accounting fees for extra
database database modules and

Web server P III 800 Current config NA

hardware 256MB RAM OK
18GB storage

Web server Sun Solaris 8 Current config NA


Web server None. Server Custom option $1,500 to

interface to currently runs available from upgrade to
SQL Server Apache web vendor for iPlanet
server Solaris/iPlanet

Client hardware Varied Current hard- NA

ware OK for
new software

Client software Windows 98 98 Certified NA

by Vendor

Copyright ©2002 SYBEX, Inc., Alameda, CA

Plan the Installation 47

Besides these costs, a number of other items must be considered and

worked into the final cost of the upgrade. These include:
 Software training for users, help desk staff, and administrators
 Downtime caused by the upgrade
 Reduced initial productivity due to user unfamiliarity with the new
database client
Training is a commonly neglected part of an upgrade process, but proper
training of users prior to the upgrade makes for a more productive—and less
chaotic—environment in the days following the upgrade. Because of this,
you should try to work training dollars into the budget early on, and make
sure that those dollars stay in the final budget!

All too often, training funds are allocated as part of the budget for a project,
but are then among the first things that managers cut out if the project is
running over budget. Fight to keep these dollars, because trained users are
happier, less confused users, and happier users make for happier administra-
tors. Remember that it won’t be the manager going desk-to-desk to explain
how the new software does this or that differently.

Once you have a good idea of what needs to be done, and how much it
should cost, you need to plan out this last cost element—as measured in
time, not money. The implementation of a large project like ours will require
a good deal of human effort. This includes administrators testing and
implementing the new configuration and help desk people dealing with addi-
tional support needs and any downtime users experience when the upgrade
process is underway. It also includes time spent in training sessions and, if
the project is complex, might include allocating funds for a consultant to
assist in crucial phases of the upgrade.
The term consultant covers a broad range of job descriptions. Some
consultants work for a particular company, and are highly specialized.
It is likely, for instance, that the software company that makes our new
SQL-based accounting package has employees who know the app well and
are expert resources on issues regarding that particular software. Other
consultants roam the landscape.
Either of these types of consultants can offer an excellent way of bringing
additional expertise onto your project. Hiring a consultant is expensive, of
course, but an experienced engineer might help you avoid problems, help

Copyright ©2002 SYBEX, Inc., Alameda, CA

48 Chapter 2  Installation

optimize the solution, and in the end save you time and money. For this
project, we will be bringing in a consultant from the software company to
review our plan and will be bringing that consultant back for the weekend of
the actual migration. To be safe, we have budgeted for four days, though
only three should be needed. As it is likely that the total bill for this assistance
will be $1,000–$1,500 per day, the consultant will be kept for the fourth day
only if absolutely needed.
The final estimate for the upgrade comes in at around $100,000 when
training, consulting, hardware, and software are all added up. This will
likely cause some serious questions to be asked about trimming costs. If
you need to reduce the cost of the project, try to cut a bit from everywhere
rather than just cutting out the consultant, the training program, or the web
server upgrade. If the company wants an $80,000 upgrade, find a way to
scale back evenly.

In the real world, there are times when the decision about what to cut is not
left up to the engineers. In these cases, you sometimes just have to deal with
the cuts where they occur, and do your best to minimize their impact. If man-
agement makes unwise cuts, that can mean that if you don’t deal with the
potential problems then, you look bad, and your users suffer. The manager
who caused this will probably be too busy golfing to even notice that there is
anything wrong.

Setting a Timeframe for Completion

Time and money. That is what it all comes down to—what does it cost, and
when can we have it? Now that you have a good idea of what resources you
will need, it is time to plan exactly how much time you will need to imple-
ment the plan. Keep in mind a few basic rules here:
1. Prioritize what items need to be started earliest.

2. Decide which tasks are dependent on other tasks.

3. Set incremental delivery dates for parts of the project.

4. Set aside time for problems.

When forecasting how long things will take, it often helps to put your-
self in the shoes of a consultant bidding a job. When a consultant bids on a

Copyright ©2002 SYBEX, Inc., Alameda, CA

Plan the Installation 49

project as a “job,” what that consultant is saying is, “I will do this project in
this many hours for this amount of money.” Because the work to be done
and the amount to be paid are fixed, the consultant must have a good idea
of how much time the project will take. If she estimates too high, her fee will
be too high, and someone else will get the bid. If she bids too low, she will get
the job but may spend a week doing a project bid out at a cost of two days.
That’s three days of free labor, which hurts.
Even if you are on salary, much the same process comes into play. If you
are too aggressive with the schedule, you might say, “We can do all that in
one day. No problem.” Of course you won’t actually be able to do that,
and you will end up backtracking, which can cause serious problems if other
people are planning off your schedule. Similarly, trying to be too safe can
make it seem like you are not interested in the project or not confident in
your ability to complete it.

Project Managers
All of this may seem confusing. You may even be thinking that you have no
interest in budgets or planning, and that isn’t what you got into computers
to do. If you are lucky, there is someone wandering around your office with
a title like “Project Manager” who deals with scheduling, budgeting, and all
the rest. If so, let him deal with all of this. Believe it or not, he might have
chosen to make that his life’s work. He will probably ask you a few ques-
tions, more than likely will require a list of project requirements as shown
above, and will later return with a spreadsheet that details the entire project
plan. Oh, and this probably goes without saying, but be nice to him—your
fate is in his hands!
If you don’t have this option, though, the following resources might come
in handy when planning the big upgrade:
 From Serf to Surfer: Becoming a Network Consultant, by Matthew
Strebe, Marc Bragg, and Steven Klovanish (Sybex, 2000). This book
deals with issues related to working as a consultant, such as planning,
pricing, and all the other stuff we have been talking about.
 Mastering Microsoft Project 2000, by Gini Courter and Annette
Marquis (Sybex, 2000). Project 2000 is reputed to be great software
for project management, and this book deals with how to use the
software and the management logic behind the process.

Copyright ©2002 SYBEX, Inc., Alameda, CA

50 Chapter 2  Installation

Verify the Installation Plan

Verification of an installation plan simply involves checking to make
sure that what you have decided to do will actually work. In order to do this,
a number of resources must be considered.
In the following sections, each of these will be examined briefly, along
with recommendations for success.

Playing Politics
As you spend more time working as an engineer, you will learn (or you may
have learned already) that the most complex aspects of computer systems are
the people who use them.
Because of this, one step you will need to take is to ensure that your
project plan is acceptable to all of the various groups and factions within
the organization. Without a doubt, the key element to surviving this part
of the project is this:
 Get a sponsor!
If the project is going to get done, you are going to need someone who
has the power to get things done. Try to identify a department or group of
users who will specifically gain from the project, and co-opt them into the
process. Going back to our example project, it seems that the accounting
department manager, or maybe even the chief financial officer, might be a
good place to look.
The reason you want to specifically target them, though, is that this
engages them personally in the success or failure of your project. If problems
come up with the budget, time conflicts, or people getting cold feet about
moving to the new system, having these people with you can make all the dif-
ference. They will be able to manipulate the system in ways we cannot even
imagine, find money in places we would never think to look, and enforce
acceptance on the troops.
Moreover, having a sponsor lends authority to the project, and often just
the sponsor’s name on a proposal will quell dissent and speed authorization.
Crazy, but true.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Verify the Installation Plan 51

Even if you do have someone to help calm the rough waters, it is also
important to continually keep people up to date on what you are planning,
and how it will affect them. Find out what concerns people have about
the project, and try to implement steps into your rollout that protect against
problems that will have a significant impact on user acceptance of the

For more information on IT project planning, consider reading the IT Project+

Study Guide, by Bill Heldman (Sybex, 2002). It covers everything you’ll need
to know—sponsors, budgets, and documentation—about managing complex
IT projects. Look for it on

Verify Hardware Compatibility with Operating System

Enough with the warm fuzzies. Our project plan has finally been approved,
the budget is set, and it is time to actually buy some new hardware and put
our plan into effect. Verification of the hardware you are planning to buy
is a critical part of this. There are few things more embarrassing than buying
expensive hardware that is not supported by the server OS or other software
you are depending on. Because of this, it is best to stick to well known
vendors, and check to see that all hardware you have has drivers specifically
written for it in your chosen OS.
The process of checking hardware/OS/application compatibility can take
you to a number of places. These include:
 The OS vendor’s website. Microsoft, for instance, uses a document
called an HCL (Hardware Compatibility List) to enumerate all hard-
ware that has been certified for use with a particular OS.
 The hardware vendor’s site. Many server solutions specifically tout
the fact that they are optimized for use with a particular OS. Some
even sell the server OS packaged with the hardware, which virtually
guarantees that the two work together.
 The application vendor’s site. This is trickier. We might want to call the
vendor of our accounting package, for instance, and find out whether
there are any known issues with certain types of hardware or software.
 Newsgroups or user forums. There are users out there working with
nearly every type of configuration possible. It is likely that at least one

Copyright ©2002 SYBEX, Inc., Alameda, CA

52 Chapter 2  Installation

of them has already tried what you are planning to try, and can give
you information on how it works.


A newsgroup is a collection of articles posted on the Internet. They are

categorized by topic, subtopic, and replies to the topics. Newsgroups can be
helpful resources but the information posted can be inaccurate. The infor-
mation and answers to questions can be posted by anyone on the Internet
who has an opinion about the topic. Always remember to verify any solution
located on a newsgroup for accuracy.

You will see all of these again, especially when we look at the objective
dealing with troubleshooting. For now, though, just remember that you
should check for compatibility when you plan, double-check before you
actually buy, and watch for problems when you implement.

Most software companies will allow you to test their software for compatibil-
ity using a trial version that has limited functionality or timed deactivation.
Certain hardware vendors also let you test out their devices, but this is gen-
erally only an option if you are considering buying a large quantity. Still, it
never hurts to ask.

Once you are certain that all the pieces of your project will fit together, it
is time to order your hardware and software, and start implementing your
project plan.

Verify Power Sources, Space, and UPS

and Network Availability
For our project, one server will be replacing another. We will therefore want
to verify that the new server will be able to fit into the location the old
one occupied. If it does not, or if you have decided to place the server in a
different location, than you should find that location early on, during the
budgeting/purchasing process.
If the server will be going into a new space, you may need to buy a server
rack, UPS (uninterruptible power supply), KVM (keyboard/video/mouse

Copyright ©2002 SYBEX, Inc., Alameda, CA

Install Hardware Using ESD Best Practices 53

switch, and monitor for the server. You may also have to budget for having
network cabling, routers, etc. if this is a new server location.

Once you have found space for your server, and identified UPS, KVM, and net-
work hookups, it is important to protect them from being stolen by someone
else. Months often pass between planning and implementation, and server
room space is always at a premium. Find some way of identifying the outlets,
network ports, and KVM ports that you will be using!

Verify That All Correct Components

and Cables Have Been Delivered
Once your hardware starts arriving, you will want to create a checklist of
some sort, so that you can keep track of what has arrived and what is still
needed. Checklists can assist tremendously with organizing and keeping
track of components as they arrive. Checklists also will help keep you on
task with the installation. Depending on the nature of the project, checklists
will vary, but they should contain detailed information relevant to the
project. If the project involves 15 servers, a couple of hundred workstations,
and some network devices, a list is utterly essential.
You will also have to remove plastic clips holding various movable
parts in place. The power supplies, for instance, are generally secured for
shipping, and the server cannot power up until you have removed their
retaining pieces.

When setting up the server, make sure you have some room. These are not
small machines, and they generally come in extremely large, well-packed
boxes. Having someone else to help lift is also a good idea.

Install Hardware Using ESD Best Practices

At this point you have a server sitting in front of you. It has no
operating system, no specialized hardware, and likely not even any hard
disks. Before you can put this new beast into production, you will need to

Copyright ©2002 SYBEX, Inc., Alameda, CA

54 Chapter 2  Installation

finish adding components, move the server into its new home in the server
room, and connect all necessary cabling.
Most servers come pre-assembled, in that their motherboards, power
supplies, and other standard components are all in place when you get them.
Most also have built-in or included network and video cards. Even so, it
is important to go through the server and check that cables are tightened
properly, cards are properly seated, and everything is in order.
It is interesting that CompTIA has chosen to mention these elements
under its first objective, “Conduct pre-installation planning activities,”
because they are referenced again in their own topics later. As such, this
book will follow a similar tack. Chapters 3 through 6 will talk extensively
about the different types of hardware available to you, and how to determine
what is the best option for your needs. In this chapter, we will concentrate
instead on just putting things together. That process is actually rather
straightforward, and includes the following steps:
 Add additional RAM, processors, or other resources to the server
as needed.
 Insert hard disks, extra CD drives, or backup drives.
 Mount the server into its rack.
 Cut and crimp cabling to the server.
 Connect the server to a UPS.
 Verify SCSI settings.
 Install external devices.
As you are working to put the machine together, be certain to take a
few general precautions. First off, make certain that you do not have power
active to the server when you are installing devices. Obviously you don’t
want to try and do this while the server is on, but even just having active
current going to the machine can pose a danger to the server—and to you.
Also, using an ESD static strap to discharge static electricity is an excellent
idea. ESD, or electrostatic discharge, is the leading reason for component
failure during installation. A static strap is a wrist strap that connects between
your wrist and a ground. The strap contains a resistor that will slowly drain
any static charges out of your body and away from the computer. ESD and

Copyright ©2002 SYBEX, Inc., Alameda, CA

Install Hardware Using ESD Best Practices 55

ESD static wrist straps are covered in detail in Chapter 15, “Disaster
Recovery.” The proper use of a static strap is shown in Figure 2.1.

FIGURE 2.1 Using a static strap

This is the sort of thing that can cause problems for experienced server
administrators. Many administrators have developed their own installation
habits over time, and may no longer use a static strap when installing. Just
remember that for the test you should do things by the book.

Installing RAM or Processors

Many servers come with processors and RAM pre-installed, but this is not
always the case. Also, any expansion kits you have purchased will need to be
added in. Installing RAM and a processor (or processors) into a server is very
similar to adding these components to any standard PC. RAM especially
is almost identical in the way it is used. Although server RAM is generally
higher quality and more expensive, it is a difference of function, not form.
Processors and motherboards will be discussed in Chapter 3, “Mother-
boards and Processors.”
With processors, a few differences must be taken into account. First off,
servers often employ expansion cards to make room for new processors. A
quad-processor-capable server, for instance, likely does not simply have four
sockets on the motherboard. Moreover, servers generally require an addi-
tional piece of hardware, called a VRM (voltage regulator module), for each
processor installed. A VRM installs on a motherboard to regulate the voltage
fed to the microprocessor. Forgetting this can seriously damage the proces-
sor and possibly the server. Figure 2.2 shows a VRM.

Copyright ©2002 SYBEX, Inc., Alameda, CA

56 Chapter 2  Installation

FIGURE 2.2 A server VRM

Make certain that the RAM and processor are seated properly, and
that the fan and heat sink for each processor are properly attached. Also
assure that the fan is plugged in.

As processors have gotten faster, the amount of heat they generate has
increased. Because of that, the processor fan and heat sink are critical to the
health of your system. Make sure that the processor fan is working by briefly
powering up the computer after installing the fan. You don’t need disk drives
or any of that—just make sure the blades are spinning properly, and power
back down.

Installing Hard Disks

One of the primary characteristics of a server-class machine is the presence of
a large drive bay, often strung off of a RAID controller. Chapter 4 will deal
with the types of drives available, and how they need to be configured. When
putting the server together, the only part of this you need to worry about
is that the drives are compatible with the server and are installed properly.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Install Hardware Using ESD Best Practices 57

Although it is possible to install drives into internal bays, many servers use
external bays into which drives are inserted. In such cases, the drives are
placed into a protective shell called a tray. This tray then fits into the bay on
the server.
In order to ensure that drives are working, you should check the inter-
nal cabling leading to the drive bay, verify that the drive trays you have
are correct for the server, and make sure that the drives themselves are
properly secured within their trays. You will also need to carefully insert
each of the trays into its bay according to vendor instructions, as in
Figure 2.3.

FIGURE 2.3 Inserting a drive into its bay

Remember to purchase the trays! Each drive purchased will need its own
drive tray, and these do not come with the server. It’s amazing how many
installations are held up waiting for trays that no one thought about until
it was too late.

Copyright ©2002 SYBEX, Inc., Alameda, CA

58 Chapter 2  Installation

Server Form Factors

There are two primary form factors for server machines—tower or rack
mount. The form factor defines the type of case that the server is housed in,
and which type you have will make a significant difference as to what steps
you take when moving the computer into its place in the server room.
Let’s look at each of these:
Tower Form Factor The tower form allows servers to be freestanding, so
they resemble normal clone PC towers. These were the most common type
of server in 1980s and early 90s. Tower servers can be nice because they
are easy to place and require no special equipment. Simply set up a table—
not some folding table, but a solid desk or workspace—and plug in the
server as you would a desktop PC. A tower server is shown in Figure 2.4.

FIGURE 2.4 An IBM tower server

Rack Form Factor A rack form server (see Figure 2.5) and a tower
server can be identical in their components or their capabilities. What is
different about them, though, is that they are specifically designed to be
placed inside space-optimizing storage cases. These server racks allow
servers and other network components to be stacked on top of each other,
making it easy to store them and also facilitating keyboard and monitor
sharing and other conveniences.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Install Hardware Using ESD Best Practices 59

FIGURE 2.5 An IBM rack server

For a look at different types of servers and a chance to go under the hood and
look around, check out Dell’s interactive 3-D demos of selected servers. Go to to
see the PowerEdge 6450, but other models can be checked out as well. Just hit
the View 3D Demo button and wait for it to install the applet.

Mount the Rack Installation

Chapter 13 on server room procedures will cover the task of how to install
racks and what the best practices are for their use. Here we will only note
that rack-mounting a server generally is a relatively easy task, but also is a
two-person job.
The first part of the install is the mounting of the rails on which the server
will sit. These are important, because if the server needs RAM added, or
some other work done, the server can slide out on the rails, rather than
having to be removed completely. After that, corresponding drive rails are
attached to the server itself and the beast can be lifted into the rack.

Many companies are now offering rack-optimized servers that are smaller and
easier to handle. They are also, of course, more expensive and more difficult
to expand, and they generally lack the drive and processor capacity of larger
rack servers or towers.

Some servers fit nicely into a standard rack, while other companies—
Compaq and Sun come quickly to mind—build their servers to their own
specs, and you usually need to buy a rack from them that is specifically

Copyright ©2002 SYBEX, Inc., Alameda, CA

60 Chapter 2  Installation

designed for their servers. The problem is generally not the width of the case,
as all cases have standardized on a 19 IN. width. Rather, the issue is that
these servers are so heavy and deep that they will topple any case that does
not have proper support.

Install External Devices

Once the server has been secured in the rack, you must attach all of the
needed cabling. Depending on the network type, the server’s network con-
nection might require installing a patch cable for each network card in the
server, or addition of the server to the network through another means.
Networking will be covered in depth in Chapter 6, “Networking.” Other
cables to be attached at this point include power cords and, if more than
one server is being used, peripheral cables leading to the KVM. Otherwise,
the keyboard, video, and mouse will be attached directly to the server.

KVM Switch
Generally just called a KVM (keyboard/video/mouse), this switch is a
component that allows multiple computers to be controlled using a single
set of input-output devices. One workspace can be set up near the rack, and
the KVM allows users to switch back and forth between servers.

One limitation of a KVM is that only one machine at a time can be managed.
To manage servers simultaneously, separate inputs and outputs must be

The KVM is able to work with multiple machines so well because it is a

powered device, and can maintain its connection to all servers, even those
that are not actively being viewed. KVM switches are often mounted
directly into the rack, or a neighboring rack, and generally come in 4-, 8-,
and 16-device versions. Larger ones are also available. KVMs generally run
$200–$600. Not terribly expensive, but also not something you can buy out
of petty cash. Well worth the money, though, and you should make certain
to budget in the cost of this device if you need it.
Switchbox For those on a budget, a standard monitor switchbox is also
available. These are far simpler, and do not have their own power source.
Because of this, they do not maintain an active connection to all devices.
Switchboxes work well for monitors, but not for keyboards and mice.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Install Hardware Using ESD Best Practices 61

Every server should have a UPS (uninterruptible power supply) serving it.
This device ensures that if the server room loses power, the server will have
time to properly shut down, rather than just shutting off. Chapter 13 deals
more with how to plan for UPS and KVM coverage in your environment. For
now, just make sure that a connection to the UPS is available, and that it is
plugged in.

Backup Drive
Backup will be covered in more detail in Chapter 14, but let’s take an
overview approach here to get your feet wet, so to speak. As you are setting
up the new server you should make sure that all of the drive space you are
adding to the network can be backed up by your existing backup strategy—
you will do this with a backup drive. A backup drive is nothing more than
a device that creates a copy of your data on a removable device such as a
magnetic tape or an optical disk (CD or DVD). If your current backup strat-
egy is not adequate, you may need to purchase another backup drive and add
it to the new server.
Backup drives generally will need a shelf in one of the server racks, and the
drive should be easily accessible, making the changing of tapes as convenient
as possible.

Backup can be done either locally or remotely. It is generally a good idea to

put the backup drive onto one of the servers that has the heaviest backup
need. For instance, if one server has 60GB of drive space, and another has
10GB, putting the backup drive on the server with the larger drive array will
save network traffic and make backup complete faster.

Backup drives usually run off a SCSI controller, although IDE/EIDE and
even floppy controller–based backup drives are available. Because of this,
the installation of the backup drive dovetails nicely with our next item:
checking the SCSI settings.

Verify SCSI ID Configuration and Termination

Most servers use a SCSI (small computer system interface) as a means of
interfacing items such as hard disks, tape drives, and optical devices with the

Copyright ©2002 SYBEX, Inc., Alameda, CA

62 Chapter 2  Installation

server. With a SCSI connection a careful configuration must be done in order

for the SCSI chain to operate as expected. Once all SCSI components are
attached, take a bit of time to go over the connections, and verify the two
critical configuration parameters of SCSI disk systems:

Setting and Checking IDs

Chapter 4 deals extensively with SCSI disks and the SCSI architecture, and
so getting into this here would be overkill. The key thing, though, is that
probably the most common reason why a server does not boot properly is
because there is a device set to the wrong ID. Remember that not only do
SCSI hard disks have to be checked, but also SCSI CD-ROM drives and
any external SCSI devices. Each SCSI device must have a unique SCSI ID

SCSI termination must occur at the beginning and end of the SCSI chain.
Remember that SCSI devices can be external or internal. Proper termination
also includes using the appropriate terminator for the SCSI type that you
are using.

Rumor has it that there is an unnaturally large concentration of SCSI questions

on the Server+ Exam. Read Chapter 4 extremely carefully, and if possible
spend some time doing outside research on the SCSI architecture as well!

Cut and Crimp Network Cabling

Once you have the server and its accessories safe in their place, it is time
to clean up after yourself a bit. Servers always have a great deal of cabling
running into and around them. An important safety step is to get them off
of the floor and out of the way.
Also, if you have had to run new network cabling for the server, it is likely
that this cabling will need to be cut and crimped for use. The process of deal-
ing with making and managing cabling will be dealt with in Chapter 6.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Verify Power-On Via Power-On Sequence 63

Verify Power-On Via Power-On Sequence

If you looked at the CompTIA objectives, you would think that an
administrator should follow these steps when installing a new server:
1. Take the server out of the box.
2. Add hardware as needed.

3. Mount it in the rack.

4. Turn it on and test it.

5. Install software.

There is nothing more depressing than hefting a new 200-pound server

into a rack only to discover that it has a bad system board and needs to be
taken out for servicing or return.

I want to discuss the steps I take when installing a new server. These are
not the steps advocated by the exam objectives; however, in my day-to-day
real life application, they work. So, I advise following the previous steps listed
for the purposes of the taking the exam but consider the steps I’m listing here
as a real-world reference.

Always, always, always test equipment on the bench to make sure it is

working before putting it into the rack.
Because of this, the order of these steps should be like this:
1. Take the server out of the box.

2. Add hardware as needed.

3. Turn it on and test it.

4. Install software.

5. Mount it in the rack.

6. Turn it on and test it again.

If your staging area has a network connection to the server room, it may
even be best to finish the entire configuration—operating system, applica-
tions, everything—on the workbench. There is generally no real advantage

Copyright ©2002 SYBEX, Inc., Alameda, CA

64 Chapter 2  Installation

to hurrying the machine into the server room, and the disadvantages include
sitting in the cold and having to enter the key code for the server room 20+
times a day while testing the system.
Regardless of when you decide to do the power-on test for your server, it
is the first real test of your skills. Hopefully all the hardware is compatible
and has been configured properly. The server should come up with BIOS
readings about the RAM and drive configuration, and it is likely that on first
boot you will be required to enter the BIOS configuration utility, so that
boards can be configured and hardware levels recorded.

I n this chapter, you learned about how to plan a project, and about
the different types of server roles. You also were given a quick overview
of the technologies and tasks that will be considered in the next few chapters.
If you have time, browsing some of the Internet sites listed in this chapter
can be invaluable in helping you to learn about the state of the industry.
They will also give you some idea of what is similar, and what is different,
among the companies producing servers for PC networks.

Exam Essentials
Know how to plan an install. Know how to define a project goal,
examine current configurations, budget a project, and set a timeline
for completion.
Know how to set a time frame for completion of a project. This
includes prioritizing, deciding which tasks are independent, setting
incremental delivery times, and setting time aside for dealing with
Know the three elements in verifying an installation plan. This includes
people, hardware, and software.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Key Terms 65

Know what the HCL is. Microsoft uses a Hardware Compatibility List
to ensure that hardware will run as expected with their operating systems.
Know what a UPS is. A UPS or uninterruptible power supply allows
you to run a server for a short period of time in the event of an AC power
Know what KVM switches are and how they are used. A KVM (or
keyboard/video/mouse) allows you to use one monitor, keyboard, and
mouse to control several computers.
Know the term ESD. Be able to explain what ESD is and how to prevent it.
Know the server form factors. Know the common server form factors
of towers and rack mount servers.

Key Terms
B efore you take the exam, be certain you are familiar with the
following terms:

backup drive static strap

drive bay switchbox
form factor tower
HCL (Hardware Compatibility UPS (uninterruptible power
List) supply)
KVM (keyboard/video/mouse) user forums
newsgroups VRM (voltage regulator module)

Copyright ©2002 SYBEX, Inc., Alameda, CA

66 Chapter 2  Installation

Review Questions
1. Which of the following is not a step in planning an installation?

A. Defining project goals

B. Examining current configuration

C. Budgeting the project

D. Researching hardware

2. When defining project goals, is it better to be detailed or focused?

A. Detailed

B. Focused

3. What should be done before performing any install?

A. Format the hard disks.

B. Plug in the new server.

C. Perform a backup.

D. Shut down the network.

4. Which of the following is not a factor to consider when working out

the final cost of an upgrade?
A. Reduced initial productivity caused by unfamiliarity with the
upgrade components
B. Software training of users

C. Resources
D. Downtime caused by the upgrade

5. What part of the upgrade process is commonly neglected?


B. Creating a list

C. Training
D. Informing users of the upgrade

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 67

6. If costs of an upgrade must be reduced, where should the reduction

come from?
A. Training

B. Software
C. Delivery methods

D. A little from everywhere

7. When setting a timeframe for completion, which of the following is

not a basic rule?
A. Prioritize start times.

B. Decide which tasks are dependent on others.

C. Address training timelines.

D. Set aside time for problems.

8. Which of the following is not a resource to consider at the “Verify the

installation plan” stage?
A. People

B. Hardware

C. Software

D. Installation Instructions

9. What does HCL stand for?

A. Hardware Compatibility List

B. Hardware Control List

C. Hot Component List

D. High Circuit List

10. Which of the following is a trait of trial software?

A. It requires registration in order to work.

B. It commonly has limited functionality or usage time.

C. It is more expensive than the full version software.

D. It can always be obtained for free.

Copyright ©2002 SYBEX, Inc., Alameda, CA

68 Chapter 2  Installation

11. What does KVM stand for?

A. Keyboard/video/mouse

B. Keyboard/video/monitor

C. Keylock/video/monitor
D. Keyboard/video/machine

12. What does UPS stand for?

A. Unstoppable power system

B. Uninterruptible power system

C. Uninterruptible power supply

D. Under powered supply

13. What does ESD stand for?

A. Electronic safety device

B. Electrostatic discharge

C. Environmental static device

D. Electronic security device

14. What is a static strap?

A. A means of connecting you to the computer

B. A metal rod to ground the computer during operation

C. A component within the server to control static

D. A device to remove charge from your body during an install

15. What is a VRM?

A. Voltage regulator module

B. Virtual resource manager

C. Voltage resource manager

D. Virtual regulator monitor

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 69

16. What is a server form factor?

A. A model of a computer

B. A term defining the type of case that houses a server

C. A class of server
D. A component within a server

17. What is a benefit of a rack server?

A. It optimizes server space.

B. It runs cooler.

C. It can house more components.

D. It is more affordable.

18. What is a drawback of a monitor switchbox?

A. It is bulky and takes up a lot of space.

B. It only works on rack form factor servers.

C. It is not powered and doesn’t maintain an active connection.

D. It is not compatible with all operating systems.

19. What is the most common problem encountered in SCSI

A. Incorrect drivers

B. Incorrect SCSI ID
C. Incorrect termination

D. SCSI component failure

20. What is POST an acronym for?

A. Power-on safety test

B. Pre-operation system test

C. Power-on self test

D. Program-on self test

Copyright ©2002 SYBEX, Inc., Alameda, CA

70 Chapter 2  Installation

Answers to Review Questions

1. D. Researching hardware is not a specific step in planning an
2. B. When defining project goals, it is better to be as focused as
3. C. Before performing an upgrade or installation, a backup should
be performed—even if the installation is not on the computer with the
data, a clash or conflict between the data computer and the new instal-
lation can still occur.
4. C. Resources are not a factor involved in determining upgrade costs.
5. C. Commonly the training of users is an area that is overlooked both
in planning and in calculating costs.
6. D. If you must slash costs, then it should be evenly distributed from
each area to ensure that there is not one area taking a major setback
or change.
7. C. Addressing training issues should be done at a later time, not at
this stage of the installation.
8. D. At this stage of the installation plan, the installation instructions
should have been thoroughly read and understood.
9. A. The HCL or Hardware Compatibility List is a component of
Microsoft operating systems that ensures that the hardware being
installed will work with the operating system as expected.
10. B. Trial software provides a limited use or time to explore the
software and see if you like it.
11. A. A KVM (keyboard/video/mouse) box allows you to control more
than one computer through the use of one keyboard, video monitor,
and mouse.
12. C. A UPS (uninterruptible power supply) provides battery power to
a computer when there is an AC power failure.
13. B. ESD (electrostatic discharge) is the transfer of static electricity
between you and a computer component.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Review Questions 71

14. D. A static strap is used during an install. It contains a resistor that

will bleed static charges away form your body and protect computer
15. A. A VRM (voltage regulator module) is used to control the voltage
sent to the processor through the motherboard.
16. B. Form factor describes the size and shape of a device, such as a
computer case or a circuit board.
17. A. Rack-mounted servers are smaller and therefore optimize server
space as compared to a tower server.
18. C. Monitor switchboxes are not powered and will not maintain an
active connection to devices. They will not work well for keyboards
and mice.
19. B. Although termination is a major issue in SCSI installation, the
most common problems are due to incorrect SCSI ID configuration.
20. C. The POST (power-on self test) occurs at system startup and can
help with troubleshooting improper installations and hardware failures.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Hardware PART

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Motherboards and

 2.1 Check/upgrade BIOS/firmware levels (system board, RAID,

controller, hard drive, etc.)

 3.2 Add processors

 On single processor upgrade, verify compatibility
 Verify N1 stepping
 Verify speed and cache matching
 Perform BIOS upgrade
 Perform OS upgrade to support multiprocessors
 Perform upgrade checklist, including: locate/obtain latest
test drivers, OS updates, software, etc.; review FAQs,
instructions, facts and issues; test and pilot; schedule
downtime; implement ESD best practices; confirm that
upgrade has been recognized; review and baseline;
document upgrade.

 3.4 Increase memory

 Verify hardware and OS support for capacity increase
 Verify memory is on hardware/vendor compatibility list
 Verify memory compatibility (e.g., speed, brand, capacity,
 Perform upgrade checklist including: locate and obtain
latest test drivers, OS updates, software, etc.; review
FAQs, instructions, facts and issues; test and pilot; schedule
downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and
baseline; document the upgrade

Copyright ©2002 SYBEX, Inc., Alameda, CA

 Verify that server and OS recognize the added
 Perform server optimization to make use of
additional RAM

 3.5 Upgrade BIOS firmware

 Perform upgrade checklist including: locate and
obtain latest test drivers, OS updates, software, etc.;
review FAQs, instructions, facts and issues; test and
pilot; schedule downtime; implement using ESD
best practices; confirm that the upgrade has been
recognized; review and baseline; document the

 3.7 Upgrade peripheral devices, internal and external

 Verify appropriate system resources (e.g., expansion
slots, IRQ, DMA, etc.)
 Perform upgrade checklist including: locate and
obtain latest test drivers, OS updates, software, etc.;
review FAQs, instructions, facts and issues; test and
pilot; schedule downtime; implement using ESD best
practices; confirm that the upgrade has been
recognized; review and baseline; document
the upgrade

 4.5 Perform hardware verification

Copyright ©2002 SYBEX, Inc., Alameda, CA

M ost server motherboards are like those of regular PCs, and
they serve similar functions. Server motherboards, however, usually contain
a few special differences. These differences include added components, as
well as advanced support for server-specific computing, which you would
not normally see on a standard PC. Before diving into specifics, though, let’s
look first at motherboards in general.
This chapter focuses on motherboards and processors. The chapter begins
with a look at the major form factors and components of a motherboard.
Focus is also given to expansion slots and RAM because they interact closely
with the motherboard. The chapter then switches focus to processors, begin-
ning with an exploration of processor interfaces with motherboards, and
then the different available processors and their specifications. Finally
you will find an exploration into some of today’s pressing issues with
processors—cooling and overclocking.

Please see Chapters 2, 7, 9, 10, 11, and 12 for additional coverage of the “Add
Processors” objective.

The Motherboard
T he motherboard is the backbone of a computer, providing connectivity
between all the components of the computer. All computer components plug
into the motherboard in one way or another. With increased demand for
computer power, designers have had to adapt motherboards accordingly.
New processors, bus speeds, RAM types, data transfer speeds, and com-
ponents have together pushed the evolution of the motherboard forward
at a steady pace.

Copyright ©2002 SYBEX, Inc., Alameda, CA

78 Chapter 3  Motherboards and Processors

Integrated Versus Non-integrated

Motherboards are classified as either integrated or non-integrated. Inte-
grated motherboards contain built-in components that are normally
expansion features. For example, a motherboard might have video and net-
work capabilities built right into the board, so you don’t have to purchase
a separate video card and network card. The obvious downside to this
arrangement is that if the video or network component ceases to function, you
will either have to replace the entire motherboard or disable the malfunction-
ing on-board component. In the worst-case scenario, the malfunctioning on-
board component can damage other components on the motherboard, neces-
sitating replacement of the motherboard.
With a non-integrated motherboard, initial cost is higher because more
individual components need to be purchased. In case of component failure,
however, replacement of the entire motherboard can be avoided. Another
issue that arises with non-integrated motherboards is the availability of
expansion slots to support the multiple components.
In server motherboards, it is common to find integrated components such
as SCSI and RAID controllers in addition to the video and network cards
already mentioned. With large amounts of RAM installed in servers, com-
bined with the fact that there is no need for enhanced video, servers often
contain some integrated components. Component failure resulting in a
smoked motherboard is still an issue though. To support these components,
a non-integrated motherboard would require a minimum of four expansion
slots, and this is the main reason components such as video and network
cards are integrated into the motherboard of a server.

Resources within a computer usually refer to memory, IRQ, or DMA. Memory

can be either physical memory (such as RAM) or virtual memory. Virtual
memory is actually hard disk space pretending to be extra RAM. IRQs are
interrupt requests. DMA is direct memory access. A DMA allows a device
within the computer to have direct access to the computer’s memory
(RAM) without having to go through the processor.

Many motherboard manufacturers offer both integrated and non-

integrated motherboards. The specific components used in integrated
motherboards vary among manufacturers, but they commonly include
on-board video, audio, modem, and network cards. Some manufacturers
will integrate all of these components; others will provide a selection.

Copyright ©2002 SYBEX, Inc., Alameda, CA

The Motherboard 79

Form Factors
Another motherboard classification is the form factor. Essentially, form
factors define the layout of components on the actual motherboard. There
are three broad categories of form factors: AT/Baby AT, ATX, and NLX.
The AT was the original IBM form factor design, on which the processor,
memory, and expansion slots were all arranged in a straight line. This posed
a problem for full-length expansion cards because the height of the processor
interfered with proper card installation. In addition, heat dissipation from the
processor sometimes caused problems for the expansion card. The Baby AT
was a smaller version of the AT with newer, smaller components. It was a more
compact board, but had the same drawbacks as the AT. In a home PC this is
rarely an issue, but in the server world many expansion cards are full-length.
Traditionally, servers are not designed around the Baby AT form factor.
The ATX component layout is different from the AT. In the ATX form
factor, the processor and memory are arranged at a right angle to the expan-
sion slots, allowing room for the use of full-length expansion cards. In
the newer computers, the combined height of the processor, heat sink, and
cooling fan make it impossible to insert full-length cards in any other form
factor, and most new computers (including servers) are built around the
ATX form factor. New ATX motherboards also offer advanced power
management features that make them even more attractive to computer
builders. For example, ATX motherboards offer a soft shutdown option,
allowing the operating system to completely power down the computer
without the user’s having to press the power switch.
NLX has been a form factor in use with desktops for quite some time. It is
a compact form factor, often referred to as a “low-profile application.” NLX
motherboards are easily distinguished by the riser card to which the expansion
cards connect. The riser card allows from two to four expansion cards to be
plugged in. These expansion cards sit parallel to the motherboard.
Servers with this form factor offer power similar to the larger traditional
servers, but in the size of a VCR. The obvious benefit of NLX is that the bulk
of a traditional server is reduced to a space-saving smaller server. Addition-
ally, servers assembled in a rack mount case can be secured to a rack, which
can itself be secured to the floor, providing better equipment safety.
Beyond these three principal categories of form factors, some companies
have created their own motherboard layout. For the manufacturer, this
proprietary design allows for specific and custom creation of servers. For
the end user or technician, however, it can be a nightmare, often requiring
special training by the manufacturer before the custom equipment can be
serviced. There is also the possible difficulty of locating the specialty parts.

Copyright ©2002 SYBEX, Inc., Alameda, CA

80 Chapter 3  Motherboards and Processors

Components of a Motherboard
Regardless of the form factor, motherboards all contain similar essential
components, including processor slots, expansion busses, RAM banks, inte-
grated controllers (either IDE or SCSI), power connectors, and peripheral
connectors. It is these essential components that work together to provide
the connectivity and communication within the computer. The diagram in
Figure 3.1 is a structural overview of a typical server motherboard.

FIGURE 3.1 Dual socket 370 motherboard

A PCI expansion bus H Power supply connector

B Integrated RJ-45 network card I Floppy disk controller
C Serial port J IDE controllers
D Parallel port K Integrates SCSI controller
E PS/2 mouse and keyboard ports L CMOS battery
F Primary processor socket 370 M RAM banks
G Secondary processor socket 370 N ISA expansion bus





Copyright ©2002 SYBEX, Inc., Alameda, CA

The Motherboard 81

Let’s take a closer look at each of these components. The following

sections should provide you with just enough information so that you
feel comfortable applying your knowledge of these components in a
server environment. We’ll cover the types of expansion busses, memory
types, processor slots, and firmware.

Expansion Busses
Expansion busses provide a means of adding additional components to
your computer, such as a video card, network card, SCSI controller, RAID
controller, or others. Integrated motherboards have less need for numerous
expansion busses than non-integrated boards. In the history of computers,
eight major expansion busses have been developed, but only three of these
busses are commonly used in modern servers: AGP, PCI, and ISA.

The Accelerated Graphics Port (AGP) bus is for advanced video. Only
one expansion card, the video card, is made for an AGP port or interface.
The AGP port is easily identified by its brown color and offset alignment
(as compared with the other expansion bus slots). Motherboards contain
only one APG port.
The first AGP release was a 64-bit data bus that ran at 33MHz (a measure
of the speed of information flow). New releases of AGP include 4XAGP,
which runs at 133MHz—four times the standard 33MHz! Although numer-
ous servers on the market include AGP video, either as an expansion port or
an on-board video card, it is unlikely that a dedicated server would need the
advantage of advanced video—how often do you play graphics-intensive
games on a server? Some “gaming servers” provide connectivity for other
gaming computers, but the server is rarely used as an actual gaming machine.
The risk of corruption or configuration problems outweighs the benefits
of the AGP bus.

First released at the inception of Pentium-generation processors, Peripheral
Component Interconnect (PCI) cards are the major expansion card type
in use today. PCI is popular due to its transfer speeds (32- or 64-bit busses)
and ease of installation. PCI also supports bus mastering (a means of
allowing a device such as a hard disk to communicate directly with another
device without the input of the CPU) and speeds up to 66MHz. Installation
and configuration are dramatically easier than for earlier busses, with
resources for the card being determined by either the operating system or

Copyright ©2002 SYBEX, Inc., Alameda, CA

82 Chapter 3  Motherboards and Processors

the BIOS—commonly referred to as Plug and Play technology. To install a

PCI card, you simply open the computer case, locate a free PCI slot (they
are off-white), remove the blank cover, and install the card. Resources (IRQ
and DMA) are configured automatically. As you will learn in the next
section on ISA, PCI configuration is a huge step forward in simplifying
expansion-card installation.

Industry Standard Architecture (ISA) is the oldest of the three types of
expansion bus. This bus preceded PCI and was extremely popular in its
time. Today it has been nearly phased out, but some motherboards still
have one or two ISA slots for use with older expansion cards. Most new
motherboards, however, have no ISA slots.
This bus is 16-bit and allows transfer speeds of 8MHz, with some models
running in turbo mode at 10MHz. ISA, at times, is difficult to configure
because it requires jumpers and/or DIP switches to be manually set. The
technician must be aware of which resources are in use and which are
available. Should the technician configure the new expansion card with
the IRQ or DMA of another device, a conflict occurs. In days gone by, it
was not uncommon to have the mouse freeze on your 486 computer when
you tried to dial up your modem because the mouse and modem were
often misconfigured with the same resources. The ramifications of this
particular conflict were minimal, but more serious situations can occur—
for example, conflict with a hard disk controller.
Table 3.1 lists the IRQ (interrupt request lines) usage for the ISA bus.
Table 3.2 lists the DMA (Direct Memory Access) assignments.

TABLE 3.1 Default IRQ Assignments for the ISA Bus

IRQ Default Assignment

0 System timer

1 Keyboard

2 Cascade to IRQ 9

3 COM 2 and 4

4 COM 1 and 3

5 LPT2

Copyright ©2002 SYBEX, Inc., Alameda, CA

The Motherboard 83

TABLE 3.1 Default IRQ Assignments for the ISA Bus (continued)

IRQ Default Assignment

6 Floppy controller

7 LPT1

8 Real-time clock

9 Cascade to IRQ 2

10 Available

11 Available

12 Bus mouse

13 Math coprocessor

14 Hard disk controller

15 Available

TABLE 3.2 Default DMA Assignments for the ISA Bus

DMA Default Assignment

0 Available

1 Available

2 Floppy controller

3 Available

4 Second DMA controller

5 Available

6 Available

7 Available

Copyright ©2002 SYBEX, Inc., Alameda, CA

84 Chapter 3  Motherboards and Processors

As shown in Tables 3.1 and 3.2, available resources are limited. Normally
IRQ 5 is available because it is rare to have and use both LPT ports. It is
also rare to use a bus mouse today. Normally a PS/2 or USB mouse is used
today, leaving IRQ 12 as a free resource. This leaves five IRQs available for
expansion cards. Typically, IRQ 5 is used for a sound card and IRQ 10 for a
network card. This is not a computer law, but rather an unwritten rule that
technicians generally follow. The sound card can be configured with any
available IRQ, but many programs take for granted that the sound card is on
IRQ 5. That leaves three open IRQs for other devices such as a network card.

Mixed Environment Expansion Busses

Most computers in use today are not the latest, greatest technology.
Many motherboards contain several different expansion busses. Brand
new motherboards will contain an AGP and several PCIs. Motherboards
that are a couple of years or more old will contain an AGP, PCI, and most
likely two or three ISA. The question is, what expansion cards do you pur-
chase for which bus? Obviously, if you have an AGP bus, then you should
try to purchase an AGP video card. This not only takes advantage of the
capabilities of the AGP, but also allows another expansion card to use the
PCI slot that the video card may have taken.

What do you do, though, if you have to choose devices to fit in PCI and ISA
busses? What if you need to install a network card and a sound card but
have just one PCI and one ISA bus left? How do you decide which expansion
card to purchase for which bus? The rule is to select the card that will be
under the most stress for the fastest bus. This will ensure that the faster
transfer rate of the bus will be put to good use. Odds are, if your computer
is to be networked, then the network card would be under more stress, so
it should be installed in the PCI bus. The sound card, although available in
a PCI format, would be the better choice for the ISA bus.

Inside the server are several different forms of memory. Each plays a
significant role in typical operation of the server. Memory in any form is
the means of storing information, either temporarily, semi-permanently,
or permanently.

Copyright ©2002 SYBEX, Inc., Alameda, CA

The Motherboard 85

Random access memory (RAM) is the most common memory within a
computer. RAM is physical memory, a collection of chips on a small
circuit board that attaches to the motherboard via a slot called a bank.
Most motherboards contain several banks for RAM installation. RAM
is volatile memory. When power to the computer is lost, the information
stored in RAM will be lost.
RAM has seen constant evolution over time. Early forms of RAM were
known as static RAM or SRAM. SRAM didn’t need constant refreshing
from the computer. Information was stored in a series of transistors that
made static RAM rather slow in sending and receiving information from
the processor. Newer RAM is called dynamic RAM (DRAM). DRAM
requires constant refreshing, and information is stored as electrical charges
within small capacitors. DRAM, due to its components, allows for high-
density packaging. This in turn creates RAM with larger capacities in much
smaller chips. Examples of DRAM are EDO, SDRAM, DDR SDRAM,

Extended data output RAM (EDO RAM) emerged in 1995. It provided a
performance increase of 10 to 15 percent over traditional memory. The
major downside to EDO RAM was that it had to be installed in pairs. If
you wanted 32MB of RAM, you had to install two 16MB modules. This
limited the number of available banks for RAM installation on a mother-
board, as well as options for RAM expansion. Six available banks on a
motherboard really meant three. Many motherboards of this era also had
specific sequencing for installing EDO RAM. For example, a computer
with 16MB installed could add another 16MB (two 8MB modules) or
32MB (two 16MB modules). EDO RAM is not in use today as a result
of these limiting factors.

This is the most common RAM in use today. Synchronous dynamic
RAM (SDRAM) runs at system bus speeds that translate into 66MHz,
100MHz, and 133MHz. These improved speeds over previous types of
RAM eliminated wait states between the system and RAM, which was an
issue in the past. A wait state is the time when the processor is waiting for

Copyright ©2002 SYBEX, Inc., Alameda, CA

86 Chapter 3  Motherboards and Processors

a response from the RAM. Everything is synchronized to the clock speed

(discussed in a later section).
Another improvement over EDO RAM is that SDRAM is installed one
module at a time. SDRAM is also available in large sizes on one module.
If you want 512MB of RAM, you can get it on a single module, leaving the
other banks available for future RAM upgrades.

You can mix higher-bus-speed SDRAM in a slower-bus-speed motherboard;

for instance, 133MHz SDRAM will work in a 100MHz-bus motherboard. The
RAM will simply run at the slower speed of the motherboard. Some mother-
boards offer a dual speed option. This allows the installer to select the bus
speed using a jumper setting on the motherboard. Remember from your
A+ Complete Study Guide that the processor, motherboard, and RAM must
match in bus speeds. Bus speeds are a measure of the speed at which the
processor, RAM, and motherboard send and receive information and instruc-
tions. These bus speeds include 66MHz, 100MHz, or more commonly the
133MHz bus speed.

Double data rate synchronous dynamic RAM (DDR SDRAM) is an
enhancement of SDRAM. DDR SDRAM provides double clock speed by
performing read and writes on both sides of the clock cycle (as opposed to
only working on one side). This translates into twice the memory executions,
and therefore increased system performance. A system with 100MHz mem-
ory bus speed will perform at an amazing 200MHz.

Direct Rambus RAM is the newest RAM available on the computer market.
It is extremely fast, with speeds up to 800MHz, and operates like DDR
SDRAM, working on both sides of the clock cycle. Rambus RAM is often used
in advanced, resource-intensive gaming systems, and is increasingly being
used in desktop computers.

Read-only memory (ROM), another important memory component, is used
to store permanent information for easy and quick retrieval. ROM chips,
much like RAM, have seen broad evolution, beginning at PROM and
moving to EPROM and EEPROM.

Copyright ©2002 SYBEX, Inc., Alameda, CA

The Motherboard 87

PROM Programmable read-only memory (PROM) was the first ROM

design. It was programmed at the factory with the necessary information
burned in as a configuration of transistor ons and offs. The PROM chip
is permanent, so making any change means replacing the PROM chip or
possibly the entire motherboard. This is very limiting in terms of oppor-
tunity to upgrade or alter the computer’s configuration.
EPROM Erasable programmable read-only memory (EPROM) stored
information as binary programming through electrical charges. Erasing
the information required shining an ultraviolet light through a special
window that wiped out all the information on the chip. This was an
improvement from the PROM but still not user friendly. You had to
disassemble your computer and then remove and replace the protective
sticker—a labor-intensive and risky task.
EEPROM Electrically erasable programmable read-only memory
(EEPROM) is the current standard for ROM. The CMOS chip, which
contains the BIOS (basic input/output system) information, is an
EEPROM chip. The chip is kept alive by the CMOS battery and,
when needed, can be flashed. Flashing is a process by which software
erases and updates the programmed information contained within
the BIOS. Unfortunately, the CMOS battery does eventually fail, and the
information saved in the BIOS program is lost. It is strongly advised that
you manually record the information found in the BIOS and then attach
it to the inside cover of the computer. Eventually, when the battery does
fail months or years later, you’ll have access to the information and can
reenter it after replacing the battery.
Many brand name computers have a proprietary BIOS; generic computers
rely on a BIOS chip made by Award, American Megatrends, or Phoenix.
Figure 3.2 illustrates a generic EPROM chip.

FIGURE 3.2 An EPROM chip

An EPROM chip

Copyright ©2002 SYBEX, Inc., Alameda, CA

88 Chapter 3  Motherboards and Processors

Cache Memory
Cache memory is located on all motherboards and operates much faster than
RAM. Cache memory stores information that is requested frequently, allowing
for faster access and response. L1 (level 1) cache memory is actually located
within the processor chip itself. L2 (level 2) cache memory is located on the
motherboard. Both L1 and L2 cache are designed to be used by the processor.
Cache memory is often found on other components as well, including
RAID cards and network cards. This cache memory provides functions to
these devices just as cache memory does for the processor: rapid storage of
and access to information.

Processor Slots
With the constant evolution of computer processors has come a change in
the way the processor connects to the motherboard. Selecting the right
processor to match a motherboard, or vice versa, is often a confusing and
difficult task. To make that task a bit easier, we have included a listing
that describes the common processor connection interfaces in detail.
Socket 1 Used with the 486 chip, often called a PGA (pin grid array) or
ZIF (zero insertion force socket). It has 169 pins and operates at 5 volts.
Socket 2 An upgrade of socket 1. Has 238 pins and runs at 5 volts. Used
for the original Pentium processors.
Socket 3 Contains 237 pins and operates at 5 or 3.3 volts, controlled by
a jumper on the motherboard. Supports all socket 2 processors as well as
the 5×86 chips.
Socket 4 With 273 pins, this socket was designed for Pentium-class
machines running at 5 volts. Beginning with the Pentium 75MHz, how-
ever, Intel dropped the voltage to 3.3 volts, so this socket had limited use.
Socket 5 Operates at 3.3 volts and has 320 pins. Supports Pentium chips
from 75MHz to 133MHz; socket 5 was replaced by socket 7.
Socket 6 Designed for the 486 at a time when the industry was moving
into the Pentium class; never really came into mainstream use.
Socket 7 The most widely used socket; contains 321 pins and operates
between 2.5 and 3.3 volts. Supports all Pentium-class chips from 75MHz
and up, and MMX chips. Also supports chips from AMD and Cyrix.
Incorporates a voltage regulator.
Socket 8 Designed primarily for the Pentium Pro chip. It has 387 pins
and operates between 3.1 and 3.3 volts.

Copyright ©2002 SYBEX, Inc., Alameda, CA

The Motherboard 89

Socket 370 A socket 7 with an additional row of pins on all four sides.
Used by Celeron processors as well as Celeron II and some Pentium III
processors (FC-PGA models).
Slot 1 A radical change in design, this was Intel’s “processor in a box.”
This boxed processor interfaces with the motherboard through what
appears to be an expansion bus, called a Slot 1. This design eliminated
the risk of bending processor pins, an all-too-common problem with
other socket interfaces. Inside the box was the same processor chip, but
preinstalled on a separate daughter card along with the L2 cache. This
was then shrouded in a heat sink and fan assembly box. Having the
processor in this format allowed for better air flow and cooling, but it
was bulky—standing on edge, it created spatial issues within the case.
Slot 2 Similar to Slot 1, but with a larger 330-contact connector slot.
This connector allowed the CPU to communicate with the motherboard
at full CPU clock speed. Slot 2 was designed for the newer Pentium chip
sets, including the Xeon processor.
Slot A Similar to Slot 1, Slot A uses a different protocol (EV6) and is
custom designed for the AMD Athlon processor. Using this protocol, the
processor-to-RAM communication can achieve speeds of 200MHz.
Socket A Using 462 pins, this socket is designed solely for the AMD chip
sets, including the Athlon and Duron processors.
Slockets A slot 1-to-socket 370 adapter. Allows a chip designed for a
socket 370 application to be used in a Slot 1 motherboard. Slockets are
not well received by many technicians. The modification between the
two interfaces is done with jumper settings and digital circuitry. Slocket
configuration is often compared to ISA expansion card configuration—
there can be serious consequences if it’s misconfigured.

In a dual processor environment (a motherboard that allows two processors

to be present), there will be a pair of processor interfaces. If these interfaces
are Slot 1, and only one processor is present, a dummy card is plugged into
the empty slot. Leave it inserted until the computer is upgraded to dual
processors. Some servers even support multiple processors. Dell, for exam-
ple, offers a single box server with four Xeon processors working together.
Some rack-mounted servers can actually link many small servers to work
together, making one large server with numerous processors.

Copyright ©2002 SYBEX, Inc., Alameda, CA

90 Chapter 3  Motherboards and Processors

Power Connectors
Electrical power interfaces with a motherboard in several ways. Main power
attaches to the motherboard through a single plug on an ATX board, or
through two smaller plugs on an AT/Baby AT. If your motherboard is of the
AT form factor, these two small plugs must be orientated correctly with
the black wires of each plug meeting in the center.
With the main power attached to the motherboard, the control of system
power is now given to the motherboard. Cooling fans, processor fan,
and startup/shutdown can all be electrically controlled and monitored by
the motherboard’s use of the electricity. For example, ATX motherboards
offer the soft shutdown feature. When you select Shut Down in the Windows
operating system, the computer will actually power down. This can be
combined with a boot initiated through a mouse movement, keyboard
hot key, or even a request through the network (called Wake up on LAN).
Most motherboards also have power connectors (2-pin or 3-pin) for cooling
fans. The fans plug directly into the motherboard where the RPMs and
airflow are controlled through the motherboard. This allows the mother-
board to maintain a consistent temperature within the computer case.
In a server environment it is very common to see more than one power
supply. This redundant power supply acts as a backup to the first one.
More time will be spent on redundant components in Chapter 5, “Fault
Tolerance and Redundancy.”

On-Board Disk Controllers

On-board disk controllers provide the interface between the disk sub-
systems and the motherboard. Typically on a motherboard there will be
one floppy disk controller. This 34-pin plug accepts a ribbon cable that
allows a floppy disk (or two) to communicate with the computer. In some
new legacy free computers, the floppy disk has been eliminated. In the
server world where items such as boot disks are often needed, it is not
advisable to get a legacy free computer to act as a server.
Hard disk controllers are also located on a motherboard. Desktop PCs
primarily use IDE controllers. Servers, although they may contain IDE
controllers, rely more often on SCSI controllers. Some higher-priced
motherboards do include an on-board SCSI controller, but traditionally
SCSI is used through a separate expansion card. With the number of
devices needed, as well as performance differences, SCSI is often the only
viable choice. Both SCSI and IDE controllers will be discussed in depth in
Chapter 4, “Storage Devices.”

Copyright ©2002 SYBEX, Inc., Alameda, CA

The Motherboard 91

Every motherboard needs a means of interfacing with the user. This is often
done through the keyboard and mouse. Once the server is up and running,
many administrators will remove the keyboard and mouse for safety and
security of the server. Access to the server is then done remotely through the
network. However, for operating system installation, a keyboard and mouse
are needed. Again, depending on your motherboard, a legacy free mother-
board may have only a few USB ports where a mouse and keyboard would
plug into. The problem is that some operating systems (Windows NT for
example) do not support USB ports. Typically, a server will have PS/2 ports
for a mouse and a keyboard. Older servers may still require a serial mouse
(which would use a DB-9 connector) and a DIN 5 keyboard connector.

Peripheral Port Connectors

Peripheral port connectors provide the interface for other devices that
will connect externally to your computer; these include printers, external
modems, and scanners. Standard peripheral ports are two DB-9 male ports
and a DB-25 female. The DB-9 male ports on a server are often used as
communication ports for an external modem or for communication to a
UPS or network connectivity device such as a switch. Again, legacy free
computers have USB ports rather than parallel or serial ports.

Jumpers and DIP switches

Depending on the type of motherboard, the setting for the processor voltage,
speed, and other system settings may be controlled by small switches that
need to be manually configured on the motherboard. Many new mother-
boards have moved away from this type of configuration for the same
reasons as the ISA bus. A simple misconfiguration can lead to serious
consequences. On a motherboard, the jumpers often control CPU voltage.
Too high of a voltage setting could permanently damage the CPU (see
Figure 3.3).

FIGURE 3.3 Jumpers





Copyright ©2002 SYBEX, Inc., Alameda, CA

92 Chapter 3  Motherboards and Processors

Firmware is defined as any software that is stored in read-only memory
(ROM, EPROM, EEPROM) and that maintains its contents when power is
removed. Inside the server are several components that will contain ROM
chips and therefore firmware. This list includes the previously discussed
CMOS chip, but also SCSI controllers and RAID controllers.
Firmware is commonly upgradable. This process normally requires
downloading the file from the manufacturer’s website. Although tempting, it
is extremely dangerous to download such files from third-party websites.
The validity and integrity of the file may be compromised. A failed firmware
update can leave you in a difficult situation. Once the correct file is down-
loaded (this will require careful matching of your hardware to the correct
download file), it can be run. The firmware update is extracted onto a floppy
disk. When the disk creation is complete, you simply reboot the computer
with the disk in and the update will occur automatically. In essence, the
firmware disk is a boot disk that runs a specific program that reprograms
the chip when the computer is booted.
Firmware updates are performed to provide new updated features and
support for the latest hardware or to repair problems with hardware. For
example, Asus Network Technologies released a firmware update for their
motherboards that will repair an issue with the soft shutdown. This firm-
ware update relates to the CMOS chip. Another example would be a firmware
update for a SCSI controller, which would provide advanced support for
new hard disk technology. No one can forget the Y2K issue—the fear that
computers would not be able to calculate the year 2000. Many systems
were repaired through a simple firmware update.
In a typical server there can be several components with firmware. It is
good practice to document each component, firmware revision number,
date, and manufacturer. Many manufacturers maintain a mailing list and
will notify you via e-mail when a new release is available. At that point,
you decide whether to update or not. The benefits of the update must be
useful to your specific application. Keep in mind that there are risks with
every update. The possibility of the device not working properly because
of conflicts with other devices, other software, or the operating system
are realistic consequences. Always perform a backup of all data as well
as a full compatibility check before performing any update to a server.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Processors 93

Processors have seen the most rapid change over the last couple of
decades. In the server environment, processors can face incredible stresses.
Selecting the right processor to meet your needs is therefore critical. Before
looking at the current processor types, you need an understanding of the
important features.
Clock speed is the main element on which most people focus when they
talk about processors. Clock speed is measured in millions of cycles per sec-
ond—megahertz (MHz). Instructions are carried out based on clock speed,
analogous to a musician playing to a metronome. Clock speed is not the only
factor in processor performance but it is a major factor. The faster the clock
speed, the faster instructions can be carried out. Latest releases of processors
have exceeded the megahertz classification and moved into the gigahertz
range. Currently processor speeds have reached 2GHz.
L1 cache, as previously mentioned, provides fast access for the processor.
Therefore processors with larger quantities of L1 cache will perform better.
Higher cache is also directly proportionate to increased price. Server proces-
sors will commonly have higher L1 cache than desktop processors.
Voltage is another consideration. Lower voltages in processors will
generate less heat, and lower heat will allow for smoother and more stable
operation. Old processors ran at 5 volts, and this generated a steady amount
of heat; however, due to the slower clock speeds, the small heat sinks were
able to handle heat dissipation. With newer processors, the heat generated
from the faster clock speeds combined with the 5 volts of direct current
electricity made the processor unstable. Most of today’s processors run
at 3.3 volts to combat this problem.

Intel Processors
It is common knowledge that Intel has had a strong hold on the computer
processor market for a long time. Today there are several other manufac-
turers of quality computer chips on the market, but Intel still has a dominant
hold on the server market. Table 3.3 defines the major Intel processor
classes, speed ranges, and specifications as seen today.

Copyright ©2002 SYBEX, Inc., Alameda, CA

94 Chapter 3  Motherboards and Processors

TABLE 3.3 Current Intel Processor Family Line

Processor Voltage Clock Speed

Pentium 3.3–5V 60MHz to 200MHz

Pentium Pro 2.5V 60MHz to 200MHz

Celeron 3.3V 266MHz to 1.2GHz

Pentium II 3.3V 233MHz to 450MHz

Pentium III 3.3V 450MHz to 1.1GHz

Pentium 4 3.3V 1.3GHz to 2GHz

P2 Xeon 3.3V 400MHz to 450MHz

P3 Xeon 3.3V 350MHz to 1,000MHz

Itanium 3.3V 733MHz to 800MHz

Any of the Intel Pentium-class processors works well within a server

environment, but Intel designed the Xeon line specifically for the high-
performance environment of workstations and servers. Xeon processors
come equipped with a larger cache, and greater flexibility while running
in a multiprocessor environment. In its latest release, Intel has touted the
Itanium processor, running with synchronous clock signaling and a 64-bit
architecture, as the ideal match for a server environment.
Intel is constantly releasing new and improved processors. The changes
most often are in the speed category, although Intel also has several new
processor lines in development. They are referred to by code name and
feature the latest in technological advancements. For more information,

AMD Processors
AMD processors, as related to servers, are rather new on the market. AMD
(Advanced Micro Devices) only recently introduced their server line chip,
the MP, to compete with the Intel server line. The MP chip is available in
a 1.2GHz and 1.0GHz clock speed and fits onto a socket A motherboard.
Standard AMD processors include the Athlon and Duron, designed to

Copyright ©2002 SYBEX, Inc., Alameda, CA

Cooling 95

compete with the Pentium and Celeron, respectively. The benefit of the AMD
processor line is the fast bus speeds discussed previously in the socket A
description. When matched with an appropriate motherboard, the AMD
processor performs efficiently. (See Table 3.4.)

TABLE 3.4 Current AMD Processor Family Line

Processor Voltage Clock Speed

Athlon MP 2.5V 1.0–1.2GHz

Athlon 3.3V 1–1.4GHz

Duron 3.3V 950MHz–1.1GHz

Alpha Chips
DEC (Digital Equipment Corporation) introduced a 64-bit processor in
1992 called an Alpha chip. Recently Compaq released a series of servers
featuring the Alpha chip, which has a superscalar design allowing the pro-
cessor to execute more than one instruction per clock cycle. It has both
an 8K data and an 8K instruction cache, and a floating-point processor.
Compared to the other processors that calculate on one side or two sides
of the clock tick, Alpha processors have a definite advantage. This has
obvious benefits in a server environment where a CPU is often required
to process a multitude of requests at a time.

B ased on the previous information on processors, it is easy to under-
stand that cooling in a server environment is critical. Regardless of the
manufacturer or model of your CPU, it will need a way to deal with heat
buildup. The faster the clock speed of the CPU, the more heat it generates,
and server processors tend to run at a high clock speed. Remember that some
servers have several CPUs. Most servers have numerous cooling fans to
assist with heat dissipation. With the dangers of overheating, fans are often
clustered together in groups with cowlings to channel air through the server.
It is also advisable to have multiple fans to allow for redundancy—should
a fan fail, another will maintain the airflow.

Copyright ©2002 SYBEX, Inc., Alameda, CA

96 Chapter 3  Motherboards and Processors

Besides cooling fans, consider other factors in maintaining proper temper-

atures within the server case. Motherboard form factors dramatically influ-
ence cooling. The ATX form factor has the processor near the power supply
fan. This allows the processor to take advantage of the cool air drawn in from
the power supply. In the AT form factor, the processor is not near this fan.
Heat sinks and processor fans are another consideration. All processors,
including dated ones (such as the 486), require at least a heat sink and cool-
ing fan. The heat sink, which rests directly on the processor, is made of alu-
minum, which dissipates heat easily. On top of the heat sink is a cooling fan.
This cooling fan assists in moving the heat that radiates off the heat sink
away from the area. After-market products are advancing dramatically to
assist in running the processor cooler. Larger heat sinks and fans that will
run at higher RPMs are available. These products are obvious favorites of
the overclockers (see “Overclocking” section, below), but should also be a
consideration for server builders.
Another option (although it can be costly) is a Peltier device (also known as
a thermoelectric [TE] module), which is a small solid-state device that func-
tions as a heat pump. This ingenious device works like a small air conditioner,
moving hot air away from the processor side and replacing it with cool air.
Peltier devices do not work well in all applications and can be difficult to
install and operate effectively.
Regardless of which form of cooling you decide on, it is mandatory that
in a server there is some form of monitoring. This monitoring watches the
temperatures and fans for any problems (fan failure, temperature threshold
reached) and then alerts the system administrator. Modern servers can send
alerts via e-mail, on-screen messages, or audible alarms. If there is no response
within a given period of time, some servers will even shut themselves down,
thus preventing any physical damage to components. Monitoring agents can
be software, hardware, or a combination of the two.
An area that is often overlooked is the server environment. With the
constant demands on a server, maintaining a consistent temperature as
well as safe environment for the computer to operate in is a must. Many
companies will create a dedicated room for the server(s). This room prevents
dust and other contaminants from disrupting the server. Many times these
server rooms have dedicated cooling facilities, such as air conditioners,
and dedicated power sources.
Another environmental concern to be aware of is the location of the
server. As you learned in the A+ Complete Study Guide, by David Groth and
Dan Newland (Sybex, 2001), it is not acceptable to have a computer in a

Copyright ©2002 SYBEX, Inc., Alameda, CA

Summary 97

location where it will face direct sunlight. Besides sunlight, other potential
concerns would be location of water pipes and electrical wires (electrical
fields). All of these could compromise server safety.

I n the world of desktop computers, overclocking is the latest rave.
Overclocking involves forcing your computer (usually the processor) to
run harder and faster than the manufacturer intended. All processors per-
form within a specific range. There is a window by which this range can be
extended and the processor made to work at the upper levels of its capabil-
ities. This is comparable to athletes who use steroids. Performance is better
but there are serious side effects. With increased processing capabilities
comes increased heat. Overclocking demands improved processor and case
cooling. Some users who have overclocked their systems go so far as to create
custom fan and heat sinks to deal with the heat issue. Before you decide to
overclock, make sure you carefully weigh the possible consequences: short-
ened life of the CPU, overheating, and damage to other components such as
the motherboard.
When related to servers, overclocking is frowned on. The processor in a
server is responsible not only for the running of the server but also shared
applications, printers, file security, authentication, Internet, and e-mail, so
the processor is working very hard already. Forcing it to work faster can lead
to catastrophic consequences. Remember, you can get along without a desk-
top for a while, but can you remain productive without your server?

T his chapter began with an exploration of motherboards. Integrated
motherboards have several common components built into the motherboard
that would otherwise be on expansion cards. These include video, audio,
modem, and network cards. Non-integrated motherboards require separate
expansion cards for each component. Benefits of integrated motherboards
include lower price, while the major drawback is the danger of component
failure that would result in replacing the entire motherboard.

Copyright ©2002 SYBEX, Inc., Alameda, CA

98 Chapter 3  Motherboards and Processors

Motherboards are available in three form factors. The less commonly

used AT boards were the original design created by IBM. Baby AT was a
smaller version of the AT created around the desktop structure. The major
problem with the Baby AT was compatibility with full-size expansion cards.
Due to the orientation of the processor, RAM, and expansion cards, the
Baby AT motherboard made it difficult if not impossible to use full-length
expansion cards. ATX is the second motherboard form factor. ATX moth-
erboards do not have the same issues as the Baby AT. Most computers today
are built around the ATX form factor. The final form factor discussed in this
chapter was NLX. This design of motherboard was created for low-profile
systems. This includes small desktops as well as rack mount systems.
We also took an in-depth look at common motherboard components.
Expansion busses provide a means of connectivity for your computer to
expand on. Today there are three major expansion busses commonly used.
These are AGP, PCI, and ISA. The AGP bus (Accelerated Graphics Port) is
designed for video cards only. PCI (Peripheral Component Interconnect)
is the major expansion bus in use today. It has a transfer speed of up to
66MHz. PCI bus resources are configured through software (including the
computer’s operating system). ISA (Industry Standard Architecture) busses
are the oldest of the three. Currently ISA busses are being phased out, but
many computers in use still have a couple of ISA expansion busses on their
motherboards. ISA busses can be difficult to configure, requiring that the
installer know about used and free resources and then be able to configure
the expansion card through jumpers.
Memory takes on several forms within a computer. RAM (random access
memory) is volatile memory used to store information temporarily while the
computer is in use. RAM has seen an evolution from Static to Dynamic
RAM. Dynamic RAM currently in use is Synchronous Dynamic RAM
(SD RAM). SD RAM is available in bus speeds of 66MHz, 100MHz,
and 133MHz. This bus speed must be matched to the motherboard and
processor. New enhanced RAM includes DDR SDRAM and Rambus RAM.
Much like RAM, ROM (read-only memory) has evolved over time, begin-
ning at PROM, then EPROM, and finally EEPROM. EEPROM chips are
the current ROM in use and can be updated (flashed).
Processor sockets were the next topic of information. Processors take on a
key role within any computer, and with changes in processors come changes
in processor interfaces with the motherboard. Other motherboard features
include on-board disk controllers, keyboard and mouse interfaces, and
peripheral port connectors (including serial and parallel interfaces).

Copyright ©2002 SYBEX, Inc., Alameda, CA

Exam Essentials 99

Intel and AMD designed processors especially for the stressful environ-
ment of a server. With these advanced processors comes the need for added
cooling. This discussion led into a section on overclocking—a risky practice
that can result in overheating.

Exam Essentials
Know the differences between integrated and non-integrated mother-
boards. Make sure to have a strong understanding of the differences,
advantages, and disadvantages of both motherboard styles.
Be able to identify the differences between motherboard form factors.
Identify AT/ Baby AT, ATX, and NLX form factors and their limitations
and benefits.
Be able to label the common components of a motherboard. This
includes expansion slots, RAM banks, processor socket, CMOS battery,
CMOS chip, power connector, and on-board controllers.
Understand the differences between common expansion busses. Be able
to identify the busses by speed, configuration, and use.
Know the IRQ and DMA resources for the ISA bus. Know what
resources are in use for what devices, as well as what resources are
Know the different types of RAM memory. Know the differences
between EDO, SDRAM, DDR RAM, and Rambus RAM. Be aware
of performance differences between the types of RAM.
Be able to explain the different types of ROM. Understand the
different levels of ROM, including PROM, EPROM, and EEPROM.
Know the different processor slots and supported processors. This
includes all sockets, slots, and slockets for Intel chips as well as the AMD
Know what firmware is and which common components have
firmware. Be able to identify a firmware chip as well as its purpose.
Know the processors that are available for use within a server. Be aware
of voltages, speeds, and common names for processors from both Intel
and AMD.

Copyright ©2002 SYBEX, Inc., Alameda, CA

100 Chapter 3  Motherboards and Processors

Key Terms
Before you take the exam, be certain you are familiar with the follow-
ing terms:

Accelerated Graphics Port (AGP) integrated

ATX non-integrated
Baby AT overclocking
cache memory Peripheral Component
Interconnect (PCI)
Direct Rambus RAM Plug and Play
double data rate synchronous programmable read-only
dynamic RAM (DDR SDRAM) memory (PROM)
electrically erasable programmable random access memory (RAM)
read-only memory (EEPROM)
erasable programmable read-only read-only memory (ROM)
memory (EPROM)
expansion busses riser card
extended data output RAM synchronous dynamic RAM
Industry Standard Architecture (ISA)

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 101

Review Questions
1. What is the difference between integrated and non-integrated
A. Integrated motherboards have built-in components normally
found on expansion cards.
B. Non-integrated motherboards are newer technology.

C. Integrated motherboards are more expensive than non-integrated.

D. Non-integrated motherboards are rarely used today.

2. Which of the following components are not normally integrated in an

integrated motherboard?
A. Network card

B. Video

C. Modem

D. SCSI controller

E. Hard disk

3. Which of the following is not an example of motherboard form




4. Which of the following is a common component of an NLX mother-

A. PCI expansion slot

B. VESA expansion slot

C. Riser card
D. ZIF processor socket

Copyright ©2002 SYBEX, Inc., Alameda, CA

102 Chapter 3  Motherboards and Processors

5. Which of the following is not a common expansion bus?




6. What does AGP stand for?

A. Accelerated Graphics Port

B. Accelerated Graphics Post

C. Advanced Graphics Port

D. Advanced Game Port

7. What is the bus speed for an AGP bus?

A. 64-bit at 33MHz

B. 32-bit at 33MHz

C. 32-bit at 66MHz

D. 64-bit at 66MHz

8. What does PCI stand for?

A. Personal Component Interface

B. Peripheral Component Interconnect

C. Personal Computer Interconnect

D. Peripheral Computer Interface

9. What is the speed of the PCI Bus?

A. 16-bit or 32-bit

B. 32-bit or 64-bit
C. 32-bit or 32-bit

D. 64-bit or 64-bit

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 103

10. How are resources configured on a PCI bus?

A. Jumpers

B. Software

C. Automatically through the BIOS

D. Manually through a boot disk

11. What is the speed of an ISA bus?

A. 8/10MHz
B. 6/8MHz

C. 10/12MHz

D. Varies, depending on motherboard

12. What is the default IRQ for LPT1?

A. 5

B. 7

C. 3

D. 9

13. What does RAM stand for?

A. Random access module

B. Repetitive access memory

C. Random access memory

D. Repetitive access module

14. What type of RAM is installed in pairs?



D. Rambus

Copyright ©2002 SYBEX, Inc., Alameda, CA

104 Chapter 3  Motherboards and Processors

15. What does the abbreviation SD RAM stand for?

A. Synchronous dynamic RAM

B. Standard digital RAM

C. Synchronous digital RAM

D. Standard dynamic RAM

16. Which type of ROM can be flashed?




17. Which type of cache is located on the processor?

A. Level 2

B. Level 3

C. Level 1

D. Level 4

18. Which socket is used for Pentium processors that run at speeds
up to 75MHz?
A. Socket 3

B. Socket 4
C. Socket 7

D. Socket 5

19. What is the most widely used socket today?

A. Socket 7
B. Socket 5

C. Socket 4

D. Socket 3

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 105

20. What is firmware?

A. Software for a hardware component

B. A program that changes the configuration of an expansion bus

C. A software update for a PCI card

D. A digital representation of a system driver

Copyright ©2002 SYBEX, Inc., Alameda, CA

106 Chapter 3  Motherboards and Processors

Answers to Review Questions

1. A. Integrated motherboards contain components that are found on
expansion cards. These include video, audio, network cards, SCSI
controllers, and modems. Integration frees up expansion slots for
other devices.
2. E. Hard disks are not integrated on motherboards.
3. C. MCA is an expansion bus. AT, ATX, and NLX are motherboard
form factors.
4. C. The riser card is used to install expansion cards within a low-
profile NLX motherboard.
5. D. IDE is actually a hard disk technology. PCI, ISA, and AGP are
expansion busses currently in use.
6. A. AGP is the Accelerated Graphics Port. It is used for video only.

7. A. The transfer speed of the AGP (first release) was 64-bit

at 33MHz.
8. B. PCI stands for Peripheral Component Interconnect. It is the most
popular expansion bus in use today.
9. B. The PCI bus can communicate at either 32-bit or 64-bit.

10. C. Resources for a PCI bus are configured automatically through the
BIOS settings.
11. A. An ISA bus can transfer at speeds of either 8 or 10MHz.
12. B. LPT1 is assigned IRQ 7 by default.

13. C. RAM stands for random access memory.

14. B. EDO RAM is installed in pairs. This was a major limitation for
this type of RAM.
15. A. SD RAM is short for synchronous dynamic RAM. This RAM
works to a synchronous clock cycle.
16. C. Electronically erasable programmable read-only memory can be
digitally erased, or flashed, as it is often called.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Review Questions 107

17. C. Level 1 cache is always located on the processor.

18. B. Intel designed the socket 4 to support Pentium-class chips up to

75MHz; in subsequent Pentium chips, they changed the voltage and
19. A. Socket 7 is considered to be the most widely used socket today.

20. A. Firmware is software that runs a hardware component—

sometimes referred to as the operating system of a hardware

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Storage Devices



 1.1 Conduct pre-installation planning activities

 Plan the installation
 Verify the installation plan
 Verify hardware compatibility with operating system
 Verify power sources, space, UPS and network availability
 Verify that all correct components and cables have been

 1.2 Install hardware using ESD best practices (boards, drives,

processors, memory, internal cable, etc.)
 Mount the rack installation
 Cut and crimp network cabling
 Install UPS
 Verify SCSI ID configuration and termination
 Install external devices (e.g., keyboards, monitors,
subsystems, modem rack, etc.)
 Verify power-on via power-on sequence

 2.4 Configure external peripherals (UPS, external drive

subsystems, etc.)

Copyright ©2002 SYBEX, Inc., Alameda, CA

 3.3 Add hard drives
 Verify that drives are the appropriate type
 Confirm termination and cabling
 For ATA/IDE drives, confirm cabling, master/slave and
potential cross-brand compatibility
 Upgrade mass storage
 Add drives to array
 Replace existing drives
 Integrate into storage solution and make it available to the
operating system
 Perform upgrade checklist, including: locate and obtain
the latest test drivers, OS upgrades, software, etc.; review
FAQs, instruction, facts and issues; test and pilot; schedule
downtime; implement ESD best practices; confirm that the
upgrade has been recognized; review and baseline;
document the upgrade.

Copyright ©2002 SYBEX, Inc., Alameda, CA

S torage devices are arguably the most important type of device
in a server. Of course, the server couldn’t run without a motherboard, pro-
cessor, and memory, and a server without a network card is pointless. But
when you think about what a server really does—stores files and databases
and provides other services—the hard disk holds everything that’s dear to
clients. Every time a client accesses the server, the client accesses the server’s
storage device.
There are a variety of storage devices that can be employed in a server.
They include hard disk drives, optical drives, CD-based drives, tape drives,
and others. This chapter will focus primarily on hard disk drives, since they
are the most common online storage devices used in servers. (Others, like
tape drives, are typically offline storage devices.)
The hard disk drives that we will concentrate on fall under one of two
major categories: IDE and SCSI. There are others, but these are the two
most common hard disk technologies. Before we get into the details about
these technologies, however, we need to understand how hard disks are

Physical and Logical Disks

A hard drive is a hunk of metal and plastic. Inside are platters that spin
rapidly, and heads that move back and forth along the surface to read and
write information. You can think of a hard drive much like a record player,
but one that can write as well. That’s pretty much all a hard drive is, at least
physically. But how the computer deals with the disk is a totally different
story. When discussing hard disks from the computer’s perspective, there are
two types of designations: physical disks and logical disks.

Copyright ©2002 SYBEX, Inc., Alameda, CA

112 Chapter 4  Storage Devices

When talking about a physical disk, we are referring to the actual device
located within a server, or connected to the server externally. For example,
a computer with two hard disks will have two physical disks. If you can hold
it, it’s a physical disk. Physical disks can refer to a variety of devices other
than hard disk drives, such as floppy disks, compact discs, and tape drives,
to name a few. The number of physical disks that your computer can have
is based on the interface type. As an example, SCSI technology allows for
more disks per controller than IDE.
A logical disk is not quite as straightforward as its physical counterpart.
One of the most common ways to define a logical disk is any space of a hard
disk (or other storage unit) that has its own disk letter.
As you learned while studying for your A+ exams, a hard disk can be par-
titioned, or divided into smaller sections. Each partition receives a drive let-
ter and therefore can be considered a logical disk. It is not uncommon for a
computer to physically have one hard disk but logically have two. This often
leads to considerable misunderstandings for those who are new to the com-
puter world. Many times people purchase a new computer assuming there
are two hard disks included in the system. After all, in the Windows envi-
ronment, they see two hard disks labeled C and a D. Later on they discover,
by either removing the cover themselves or having someone inform them,
that they actually have only one physical hard drive.
Windows-based systems can support up to 23 logical disks each. After
that, we run out of letters in the alphabet. Yes, the English alphabet has 26
letters, but remember that A, B, and C are reserved for two floppies and the
first hard disk. NetWare and Unix do not use drive letters. They name their
logical disks (called volumes) by referring to the logical unit by machine and
then a volume name. For example, on a NetWare server, there is always a
SYS volume. If your NetWare server is named NW1, the volume name (log-
ical disk name) would be NW1/SYS:.
In a network environment, you will frequently work with multiple logical
disks. You will often find network mapped paths, which also qualify as a
logical disk. These mapped paths are given a unique drive letter and for all
intents and purposes act as another hard disk for the end user. The difference
is that these disks do not reside on the end user’s computer but rather on the
server, or other computer, and are accessed through the network. Network
mapped paths are a logical pointer to a physical resource. This tricks the cli-
ent computer into thinking that it has another hard disk, which in actuality
is a part of a hard disk on another computer. Your computer may think it has
an H disk, and it’s right. The trick is, the information is physically located
somewhere else.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 113

Should the host computer lose connection to the network, the mapped
path will not be accessible. Figure 4.1 illustrates the view of a mapped path
through the My Computer icon in Windows 2000. Notice how the mapped
path icon for disk F is similar to that of the hard disks, but has a small
network wire connector visible below the disk picture. This small network
wire connector is there to remind you that this is a mapped drive. If the
network-mapped disk is unavailable, it will have a red X on the network
wire connector.

FIGURE 4.1 A mapped path displayed in Windows Explorer

Also on a network, you may run into multiple physical devices that form
one logical disk. A good example is a RAID 1 or RAID 5 array. There will
be at least two physical hard disks (three for RAID 5) that appear as one
logical unit.
So with all this talk of physical versus logical disks, what’s the real deal?
It seems slightly confusing. Just remember that if you can hold it, it’s a
physical disk. However, logical disks are all about how you define them.
A physical hard disk may have multiple logical disks, and at the same time
multiple physical hard disks can make up one logical disk. It all depends on
how you set it up. Last but not least, logical disks can also be physically
contained on another machine, but you have a drive letter on your machine
representing it.

Copyright ©2002 SYBEX, Inc., Alameda, CA

114 Chapter 4  Storage Devices

Storage devices interface with the computer through two possible means:
either an IDE or SCSI interface. Because the connectors are so different,
IDE devices cannot be on a SCSI controller, and vice versa. However, some
computers do have connectors for both types of disk. Depending on costs
and requirements, choosing one over another is sometimes difficult in the
desktop/workstation environment. In the server environment SCSI is the
dominant storage device because it is much more extensible, it is compatible
with a wide variety of devices, and it has evolved to operate at many different
levels and speeds. Basically, it’s faster and more expandable. However, it is
also more expensive. In the next two sections, we will explore each option,
for a means of comparison and also for configuration purposes.

IDE Technology
IDE (Integrated Drive Electronics) technology was first created as a simple
means of adding components to a computer. Today, this technology is more
commonly associated with ATA (Attachment Interface) technology. The
controller circuitry is located right on the device itself. The device is then
attached to the motherboard or expansion card with a short 40-pin ribbon
cable. Most cables are keyed so they will fit on only one way. If the cable is
not keyed, then the rule is that the red stripe on the cable connects to pin 1.
Closely examine the cable and you will see that one side of the ribbon has
a red line (although sometimes it’s blue). Also a close examination of the
device will reveal that the connector is labeled with a pin 1. If you can’t
locate pin 1, it’s always on the side of the connector closest to the power con-
nector on the hard disk. A simple jumper configuration (discussed later in
this chapter) provides a means of configuration, and the IDE device is
installed and running.
The original release of this technology supported disks up to 528MB in
size and transferred at a speed of 3.3MBps. With increased disk size ATA-2
(version 2) was released that supported several gigabytes of disk size as well
as increasing speed to 11.1MBps. This release was commonly known as
EIDE or Enhanced IDE technology. The latest release is Ultra DMA/33
(Ultra Direct Memory Access 33) IDE. It can transfer at speeds of 33MBps.

Newer versions of Ultra ATA IDE can support 66MBps and 100MBps. They are
called Ultra ATA/66 and Ultra ATA/100, respectively.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 115

Regardless of the release or version, IDE has one major limitation, which
is that it can support only two devices per controller. In the days of its incep-
tion this was not seen as problematic because hard disks were the only
devices supported. However, with the creation and widespread use of CD-
ROMs, DVDs, and CD-RWs, the need to support multiple devices grew. All
new motherboards tried to meet this demand with support for two separate
IDE channels, often referred to as IDE controllers (primary and secondary).
This would then allow for four devices (two on each channel). In today’s
world of multiple physical disks, burners, and DVD drives, IDE technology
is not the most flexible. Also, 33MBps may seem wonderfully fast, but
compared to SCSI it takes a back seat. And because of overhead associated
with the technology, you will only get about 75 percent of the theoretical
maximum transfer rate with IDE hard disks.
In the server environment speed is essential. Access time to server hard
disks when multiple requests are incoming can add up to rather long wait
periods. If the request is from an application such as a database, the possi-
bility of a time-out error is real. A time-out error happens when the program
gives up waiting for a response from the server. IDE devices have tried to deal
with wait state issues by increasing hard disk spin rates (the same holds true
for SCSI). Typical hard disks spin at 5,400 RPM (revolutions per minute,
referring to the speed at which the platters spin). New releases in IDE
hard disks include 7,200 RPM and 10,000 RPM. Theoretically the faster
the hard disk can spin the faster that the actuator arm and read/write head
can get to the data stored on the platters. This has improved performance in
hard disks but still doesn’t deal with the primary issue with IDE technology:
support for only two devices.

Configuring IDE Devices

Configuration of IDE devices, as previously mentioned, is rather simple.
Jumpers are located on each device and allow for three possible configura-
tions. The first configuration is MA, or Master setting, and it is used if
the device is the only one on the channel or if it is the primary device when
two devices are present on the same channel. The second possible configu-
ration is SL, or Slave, and it is configured only if the device is the secondary
device on a channel. Thirdly, CS, or cable select, is used if the devices can
auto-configure for master and/or slave. Jumper configurations for IDE are
illustrated here for you in Figure 4.2.

Copyright ©2002 SYBEX, Inc., Alameda, CA

116 Chapter 4  Storage Devices

FIGURE 4.2 IDE configuration

Setup Jumpers

Jumpers- Style A
ATA Hard Drive, Rear View

ATA Cable Key ATA Cable Pin 1 Jumpers

An extra spare (horizontal)

jumper may also be
present. This does not
Master Enabled Slave Enabled Cable Select affect the drives use.
(CLJ) Capacity limitation jumper.
The CLJ jumper is used together
with the Master, Slave, or Cable
Select setting.

Most administrators avoid using CS on their hard disks because it slows down
the boot process. This is because during the boot process the computer has
to scan the IDE cable to detect if devices are present and then determine which
is to be master and which slave.

Practical experience has shown technicians that certain combinations

work best. For example it is recommended that the boot hard disk be
attached to the primary IDE controller in the master position. This will assist
in faster boot. Many motherboards include one Ultra DMA controller (the
other controller is standard IDE to keep costs down) in the primary IDE

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 117

position to further assist with speed to the primary boot disk. If you have a
CD-ROM and a CD-RW it is advisable to connect both on the secondary
channel. This will help to prevent buffer under run errors that can occur
when the CD-ROM is on a different IDE channel from the CD-RW.
Once the jumpers are configured for the master and slave settings, the IDE
devices can be installed into the computer case. The final step involves enter-
ing the setup utility when the computer is booting. In your BIOS setup is an
option to manually specify what is connected to each of the master and slave
positions for both IDE controllers, or you can have the BIOS autodetect
these devices each time the computer is booted. If you choose to have the
BIOS autodetect these devices, expect the boot process to become slower.
An option that can be very helpful when installing a hard disk is the Auto
Hard Drive Detect. This utility, found within the BIOS setup program, will
scan the IDE channels and try to identify the hard disks and appropriate
settings for you. After the utility scans the disks, it will present you with its
findings for you to accept or reject. If you accept, they will be automatically
set in the BIOS.

Older BIOSes did not have the ability to autodetect hard disks. You had
to enter in configuration information manually. Even though it may be a
bit slower to have your computer autodetect the hard disks, it’s still a
valid option.

Common IDE devices used today include hard disks, tape drives, CD-
ROM, DVD, CD-RW, and internal Iomega Zip drives. Combinations of
these devices are found in every computer from laptops to servers. Each uses
the previously mentioned jumpers and master/slave settings.
With the limitation of two available devices per channel, IDE is not
common in servers. Many servers contain more than two hard disks alone,
which would leave no connectivity options for a CD-ROM drive or a backup
drive. Transfer speed is another concern. Although IDE speeds seem impres-
sive in a stand-alone computer, when placed in a server environment where
the speed is shared among many client computers, IDE struggles at times
under the demands. More common in a server is the SCSI structure for stor-
age devices. This is not to say that there is no IDE at all in servers. Server
motherboards normally contain at least one onboard IDE slot.

Copyright ©2002 SYBEX, Inc., Alameda, CA

118 Chapter 4  Storage Devices

Getting IDE Drives to Play Nice

Some years ago, I was working as a consultant for a small insurance office.
They only had four computers, and no servers. It was a simple configura-
tion, and they didn’t require a lot of maintenance.

They came to the point where they needed additional storage space in the
workstation that held their database. So, they bought an additional hard
drive, and called me to put it in.

I arrived at the office, powered the machine down, and removed the case. I
grounded myself properly (notice, good ESD safety being practiced), and
removed the existing drive. Sure enough, the drive was not jumpered as a
master. So, using the diagram on the drive itself, I set the jumpers to mas-
ter. Looking at the new drive, I jumpered it as a slave. I put both drives back
in the machine, and powered it up.

It didn’t boot. So I checked the system BIOS, and manually configured the
master and slave settings based on the drive parameters. We rebooted
the machine, and still nothing. Just as a test, we changed the jumpers
on the new drive to make it a single, changed the BIOS, and sure enough,
the drive booted. So we knew that the new drive was good. And the old
drive had been working a few minutes ago. So what could be the problem?

We set it back up again with the old drive as the master, and the new drive
as the slave, and got the same results. It didn’t want to play. Looking at the
drives again, I noticed that the old drive was a Seagate, and the new one
was a Maxtor. Puzzled, I switched the master/slave relationship of the
drives, changed the BIOS, and the system booted fine.

The moral of the story is this: sometimes when dealing with multiple
IDE drives, they may not work in a specific master/slave relationship. Try
switching them to see if that helps. It’s not necessarily a Seagate versus
Maxtor thing, but it is more common when you use drives made by differ-
ent manufacturers. So if you can, stick to using drives made by the same
company. It’s also somewhat common for older drives to not work as mas-
ters to newer drives. Again, change the master/slave relationship, and you
should be okay.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 119

SCSI (Small Computer Systems Interface) is far more robust than the IDE
structure. Unfortunately it is also far more complex in configuration and
setup, and more expensive. When you talk about SCSI devices, the discussion
is not limited to hard disks. Available SCSI devices include a broad range
of internal and external components. In the server environment, the range is
often dominated by hard disks, tape backup drives, and CD-ROM drives.
However, there are also SCSI scanners, optical devices, and others. The fol-
lowing section focuses on the fundamentals of SCSI hard disks.
The SCSI standard was put into effect in the mid-1980s and specifies a
universal, parallel, system-level interface for connecting up to eight devices
(including the controller) in a chain on a single shared cable. This grouping of
devices is called a SCSI bus. SCSI busses are extremely flexible in design. The
SCSI controller card controls the devices, so you can be confident that, as long
as the card works in the computer, then the SCSI devices will also. Therefore
SCSI devices will perform equally well in a PC, a Mac, or a Sun Microsystems
workstation, as long as the controller card itself works with the operating
system and other hardware. The SCSI controller card contains its own config-
uration as well as firmware. SCSI supports many more devices than available
in the IDE technology, and also transfers information at much faster speeds.
All SCSI configurations require termination at both ends of the chain.
If there is no termination, the signal will bounce back and forth along the
chain, causing the devices to fail. SCSI adapters have a terminator built in,
and you must supply the terminator at the other end. SCSI devices are iden-
tified by a SCSI ID number. The controller typically takes ID 7, and the
devices get 0 through 6.

SCSI Signaling Types

There are three distinct SCSI signaling methods. Careful consideration must
be given to the signaling method chosen because all devices, cables, and
adapter cards on one chain should be of the same signal type, and must be
of the same signal type if you are using HVD (High Voltage Differential).
The three signaling methods are:
SE Most SCSI devices use single-ended (SE) signaling. This method has
a maximum bus length of 1.5 meters, and uses a 50-pin narrow SCSI
connector, such as the Ribbon IDC 50 connector. SE signaling requires
termination. (See the “SCSI Termination” section below for details.)

Copyright ©2002 SYBEX, Inc., Alameda, CA

120 Chapter 4  Storage Devices

HVD High Voltage Differential (HVD) (also known as differential)

is used to provide reliable signals in a high-noise environment over a
long bus length (up to 25 meters). The HVD signaling method cannot be
mixed with other signaling methods, and requires a differential (HVD)
LVD Low Voltage Differential (LVD) supports a bus length of 12 meters
and supports downward compatibility with single-ended signaling devices.
LVD-compatible equipment is required for the Ultra SCSI standards. LVD
requires termination.

SCSI Types
SCSI technology has seen a constant and dramatic change since its incep-
tion. The first release of SCSI technology was rather awkward and limiting.
Still, the potential was clearly evident. Later releases improved on predeces-
sors in areas of speed and reliability. The following is a brief look at the
essential elements of each major SCSI release.
SCSI-1 The first true implementation of SCSI was SCSI-1, created in
1986. It had a 5MBps transfer rate and used a Centronics 50-pin cable or
a DB-25 female connector with an 8-bit bus width. SCSI-1 was based on
single-ended transmission and used passive termination. Passive termi-
nators had only resistors to terminate the bus, as opposed to active ter-
minators that have voltage regulators for added reliability. The original
release of SCSI was not without problems. While there were standards,
there was inconsistent implementation of the standards by vendors.
SCSI-1 is now obsolete. If you mix SCSI-1 devices on a bus with other SCSI
devices, performance will degrade.
SCSI-2 The goal of SCSI-2 was to improve on performance and reliabil-
ity, and to enhance features. SCSI-2 was also needed to standardize the
commands used with the technology. SCSI-2, which was backward com-
patible with SCSI-1, introduced a higher-density connector and both an
8-bit and 16-bit wide bus. The 16-bit bus was known as Wide SCSI-2.
SCSI-2 also introduced a faster speed release called Fast SCSI-2, which
used the 8-bit bus but at a speed of 10MBps. It was also possible to com-
bine the best of both Wide SCSI and Fast SCSI to get SCSI-2 Fast-Wide
(16-bit at 40MBps). SCSI-2 also used active termination, which is more
reliable than the passive termination used in SCSI-1.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 121

Wide Ultra-2 SCSI Wide Ultra-2 is a step up from SCSI-2. This release
provided LVD or HVD signaling, a 16-bit wide bus, transfer speeds of
80MBps, LVD or HVD termination, and used a 68-pin connector.
Ultra-3 SCSI Ultra-3 is the latest SCSI standard. Ultra-3, also called
Ultra SCSI, operates at a faster 20–40MBps, which was a definite
improvement over previous SCSI releases. Ultra-3, however, addressed
another problem: cable length. With SCSI-3, LVD was introduced. Low
Voltage Differential increased the possible length of the cable to 25 meters
with a possible transfer speed of 160MBps. Ultra-3 SCSI operated at a
16-bit wide bus, with LVD signaling and termination. Ultra-3 used a
68-pin connector.
Ultra 160 This release is a subset of Ultra-3. It is a parallel interface that
uses a 16-bit wide bus and LVD signaling and termination, and has a
maximum transfer speed of 160MBps. Although similar to Ultra-3, Ultra
160’s faster transfer speed and LVD addition warranted the creation of
this new SCSI category to prevent compatibility issues between device
vendors. Ultra 160 also used a 68-pin connector.
Ultra 320 SCSI Ultra 320 is the next generation of parallel SCSI inter-
face. At one point it was called SCSI Ultra-4. It is a 16-bit wide bus that
uses LVD signaling, LVD termination, a 68-pin connector, and has a
transfer speed of 320MBps.
Table 4.1 helps to clarify the various releases of SCSI and their speeds and
cable specifications.

TABLE 4.1 Common SCSI Standards

Transfer Cable
Type Bus Width Rate (MBps) Connector Length

SCSI-1 8 5 DB-25 6m

SCSI-2 8, 16 5 C-50 6m

Fast SCSI 8 10 C-50 3m

Wide SCSI 16 20 C-68 3m

Wide Ultra-2 16 80 C-68 3m

Copyright ©2002 SYBEX, Inc., Alameda, CA

122 Chapter 4  Storage Devices

TABLE 4.1 Common SCSI Standards (continued)

Transfer Cable
Type Bus Width Rate (MBps) Connector Length

Ultra SCSI 8 20 C-50 3m

Ultra 160 16 160 C-68 12m

Ultra 320 16 320 C-68 12m

SCSI Controller Card

The SCSI controller card is the heart of the SCSI system. Unlike IDE tech-
nology, where the controller circuitry is located within each IDE device, SCSI
control resides on the adapter card solely. The SCSI adapter card contains a
firmware and an EEPROM chip similar to the motherboard. It is on this chip
that the SCSI adapter card maintains the configuration information for the
individual components installed on the SCSI bus. SCSI adapter cards can
support internal or external devices. Depending on the SCSI standard imple-
mented, the connectors and supported cabling to the adapter card will vary.
Figure 4.3 illustrates a typical SCSI adapter card.

FIGURE 4.3 SCSI controller card

50-pin connector
Ultra160 LVD connector

50-pin connector

32-bit PCI

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 123

SCSI adapter cards can vary from a simple controller card that is
packaged with SCSI scanners to very expensive multichannel models.
The type of card you choose will depend on your budget as well as needed

SCSI Installation and configuration

Installing a SCSI device is substantially different from the previously dis-
cussed IDE technology. Rather than a simple master/slave jumper setting,
there are several key configurations that must be carefully selected. Each of
these configurations plays a significant role in the SCSI operation. Once the
SCSI card and devices are purchased, they need to be configured, usually
though a bank of jumpers located on the device itself. Configuration
includes setting the SCSI ID, LUN (logical unit number), and termination;
within each of these configurations, care must be taken to ensure proper
operation. These configurations will be examined in detail later in this
section. Be sure to document all settings and configurations during the
installation and configuration process to assist in future upgrades,
additions, and troubleshooting.
The following section will look at the fundamentals behind SCSI config-
uration, such as SCSI IDs, SCSI connectors and cables, and SCSI termina-
tion. Server administrators will almost definitely encounter SCSI devices
and should therefore have a basic understanding of these fundamentals.
This topic is covered in great detail on the Server+ Exam.


Each device, including the controller card, must have a unique SCSI ID num-
ber because this number is used to identify each component on the SCSI
chain. The available numbers are 0–7 on an 8-bit SCSI bus, and 0–15 on a
16-bit SCSI bus.
The boot hard disk is normally set to ID 0. This is because the controller
card starts its configuration at 0 and thus will find the boot disk first, theo-
retically resulting in a faster boot. Also, some host adapters will not boot if
the hard disk is not set to ID 0. SCSI ID numbers are configured through a
jumper setting on the device or, in the case of the SCSI adapter card, through
a software utility. Figure 4.4 shows the jumper bank on a SCSI hard disk.
Note the possible settings are for an 8-bit bus.

Copyright ©2002 SYBEX, Inc., Alameda, CA

124 Chapter 4  Storage Devices

FIGURE 4.4 Jumper configuration


Oftentimes a single SCSI device will perform multiple functions. In this

case, the device will need a means of identifying each function through an
address (similar to the unique SCSI ID). This is where the LUN (logical unit
number) steps in. A good way to grasp the LUN concept is to compare it to
a shopping mall. The building as a whole has one street address but each
shop within the mall has its own separate address. These separate addresses
are analogous to the LUN. In the SCSI-2 standards the LUN numbers can be
from 0 to 7. A practical example of this is a tape changer. The tape drive
would require one LUN number while the changer would require another.
This is one SCSI device that performs two separate functions and would
therefore require two LUN numbers.

Most SCSI cards are configured to ID 7 by default, but you might sometimes
need to reconfigure your adapter card to a different ID. You do this through
the software configuration utility.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 125

SCSI Termination
SCSI termination seems simple enough. Place a terminator at the beginning
and at the end of the SCSI chain (just as you learned while studying for your
A+ exam) and everything should work. However, the majority of configura-
tion problems with SCSI occur with termination or ID numbering. SCSI
termination can be a difficult task because of the many variables you must
First, termination must occur at the ends of the SCSI chain. Most internal
devices are terminated through the use of a jumper. Be sure to locate the cor-
rect jumper and apply it appropriately. Each manufacturer has a specific
sequence in which to properly apply or remove termination. If the SCSI
controller card is to be terminated, it is normally done through software con-
figuration. During system bootup you will see an option to enter the SCSI
configuration utility, in which adapter card termination can be enabled or
disabled. Where termination takes a much more difficult turn is when exter-
nal SCSI devices are introduced. When both internal and external devices are
present, the termination is removed from the SCSI adapter card and then
configured on the device at the end of the internal and external chains. With
external termination, a secondary device is often needed. Depending on the
type of SCSI bus, the terminator will vary. There are five basic types of SCSI
Passive Termination consists of a 220-ohm resistor that connects to the
terminator and a 330-ohm resistor that connects the signal line to the
ground. Passive termination is less expensive but can lead to issues with
line noise and dirty signals. Passive termination is not recommended for
SCSI-2 configurations. HVD SCSI, however, does use passive termination.
Active Termination was created to eliminate signal problems experi-
enced with passive termination. Active termination is based around a
voltage regulator, which reduces fluctuations. Active termination uses
a single 110-ohm resistor. SCSI-2 uses active termination.
FPT uses diode switching and biasing to fill any fluctuations between
the cable and devices. FPT (Force Perfect Termination) is a more
advanced form of active termination.
LVD is based on a form of active termination. Low Voltage Differential
(LVD) is based around the higher speeds of SCSI Ultra-2. Special LVD/SE
(single-ended) terminators can be used on busses that have both LVD
and SE devices.

Copyright ©2002 SYBEX, Inc., Alameda, CA

126 Chapter 4  Storage Devices

HVD High Voltage Differential (LVD) is based around a form of active

termination. HVD devices require HVD termination, and should not be
mixed with LVD or SE devices.
As you can now see, there is ample opportunity to create problems in ter-
mination. Each terminator must match the appropriate SCSI type. If an inap-
propriate terminator is used, the SCSI bus will not function properly and the
termination could damage devices on the SCSI bus. Considering the cost of
SCSI devices and adapter cards, care should be taken before powering on
your system. For the best information, refer to your SCSI device
After you have completed all of the configuration steps, you can install the
devices in the same way as IDE devices. The SCSI cable connects between
each device and the adapter card. A power connector also connects to each
device. Figure 4.5 illustrates the connectors at the back of a SCSI hard disk.

FIGURE 4.5 SCSI hard disk connectors

68 pin Centronics Connector Jumpers Power Connector

After all the devices are attached and installed, review each device to con-
firm proper cabling and termination. When you are sure that installation is
complete, you can start the server and enter into the SCSI utility. Depending
on the manufacturer, this can be done in several ways. Normally access is
gained during the bootup of the computer. A line of text will appear on the
screen telling you to press a specific key or sequence of keys to enter the SCSI
setup utility, where you can confirm that each device is identified by the SCSI
adapter (based on the SCSI ID and LUN) and is functioning properly.

Most SCSI cards are terminated by default. If you are connecting internal
and external SCSI devices to your chain, be aware that you will have to
disable the termination on the adapter card. Figure 4.6 is an example of an
external terminator.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 127

FIGURE 4.6 SCSI terminator

SCSI Cables
SCSI cables play an important role in the SCSI chain. SCSI cabling today
comes in many different forms. Remember that SCSI devices can be internal
or external. Internal cabling differs significantly from external cabling in
terms of durability as well as reliability. Consider that cable connectors also
vary depending on the SCSI standard being implemented, and it becomes
easy to see that there are numerous possibilities.
Internal cables follow two different forms:
 Standard ribbon cable (similar to IDE and floppy cable) is commonly
found within the server case. This cable normally is 68 wires wide (to
accommodate the 68-pin connector) but can also come in a 50-pin form.
 Newer internal cable is twisted-pair cable and looks a little like spa-
ghetti. This cable is round, not flat like ribbon cable, and has twisted
pairs of wires for each pin. The idea behind using the twisted pairs is
to reduce signal degeneration. This cabling costs more than traditional
ribbon cable, but does improve signal stability. It is often used in
longer SCSI bus implementations. Twisted-pair cable can also be
found with metal braided shielding surrounding the twisted pairs,
which protects further against signal interference.
External SCSI cables need to be more durable than internal cables. Being
exposed to the environment and environmental hazards, these cables have a
strong external sheathing to protect internal wires. Many external SCSI cables
also contain ground shielding, which protects against signal interference.

SCSI Connectors
SCSI connectors physically attach the drive to the cable. Several different
types, such as Centronics connectors, are available to meet the demands of

Copyright ©2002 SYBEX, Inc., Alameda, CA

128 Chapter 4  Storage Devices

bus width and speed. Figure 4.7 illustrates some common connectors that
you may encounter.

FIGURE 4.7 Common SCSI connectors

DB-25 Connector used with Centronics 50 Connector

SCSI-1 standards used with external SCSI devices

Internal and External 68-pin connector Internal 50-pin ribbon cable

used by most hard drives connector for SCSI-2

Very High Density Centronics 68-pin 80-pin connector used by Ultra 160
used on SCSI-3 and Ultra 2

Benefits of SCSI over IDE

Now that we have compared and contrasted both technologies in detail, you
are probably realizing that there are several reasons why SCSI implementa-
tion is preferred over IDE technology. Even though configuring the SCSI bus
can be a painful task, and the initial cost of SCSI devices is higher than that
of IDE devices, SCSI is still preferred in many applications. As a rule of
thumb, if you have a server, go with SCSI. Here is a summary of SCSI
benefits over IDE.
Speed The transfer speed of SCSI is dramatically faster than IDE. A
server will face a barrage of requests at once, and SCSI technology can
better meet this stress load.
Number of Devices Servers contain multiple hard disks, especially in
RAID configurations, where disk arrays provide redundancy (discussed in
Chapter 5, “Fault Tolerance and Redundancy”). IDE technology without
special accommodation will support only four devices. A single 8-bit SCSI
channel can support seven devices (remember, the controller needs one).

Copyright ©2002 SYBEX, Inc., Alameda, CA

Physical and Logical Disks 129

It is common now to see an adapter card with numerous SCSI channels

providing you with nearly limitless device capacity.
Controller Card All SCSI devices are controlled by the SCSI controller
card. This takes the stress away from the processor and motherboard.
Remember that in the IDE environment the devices each have their own
controller circuitry, which is controlled by the BIOS program, which is
located within the CMOS chip, which is on the motherboard. Placing this
burden on a separate card frees up resources for other uses.
Multiple SCSI Controllers Server capacity is limited only by cost. If you
need more SCSI devices and your current bus is full, simply install another
SCSI adapter card! As long as there are available PCI slots, you can install
another SCSI bus.
Internal and External Support IDE technology is limited to the space
within the computer case. SCSI on the other hand will support internal
and external devices. Many SCSI controller cards have external 68-pin
connectors for adding on external devices.
Although initially SCSI devices, cables, and adapter cards cost more than
comparable IDE devices, their speed and flexibility more than make up for it.

Hard Disk Administration

Hard disks are electronic components, and like all electronic components,
their performance will degrade over time until eventually they need to be
replaced. It is usually better to replace the disk drive before it fails rather than
after. You will know a disk is failing if performance drops off drastically
without any new strains being imposed on the device.
When replacing an existing disk, it’s a good idea to upgrade it at the same
time. Newer, faster, bigger, and cheaper disks are always coming onto the
market, so expand your storage capacity if you can. Before replacing a hard
disk, remember to shut off power to the server, and remember to practice
proper ESD safety procedures.

Disk Arrays
If you have multiple disks in a RAID configuration, you should first check
with your manufacturer’s documentation. If software is controlling the
array, it may be difficult to expand. As an example, in Windows NT and
Windows 2000, if you have RAID 1 or RAID 5, you cannot expand it

Copyright ©2002 SYBEX, Inc., Alameda, CA

130 Chapter 4  Storage Devices

without deleting the existing array and creating a new one. Of course, if
you do this, please make sure to back up (and test) your existing data first.
(See Chapter 5, “Fault Tolerance and Redundancy” for more information
about RAID.)
Many manufacturers provide external disk arrays for use with servers.
The disk arrays connect to the server through a proprietary expansion card.
These devices are often very expensive, but have some major benefits. Most
of these external storage units contain their own processor and memory,
which makes them very fast, and they do not drain excess resources from
your server. They are also very expandable. Also, most of them use hot-
swappable disks. If one fails, you will get a red indicator light next to it.
Simply pull it out and put a new one in, and the unit will integrate the
disk automatically for you. Technology is a beautiful thing.

Disk Installation Notes

Always remember that disks must be formatted for the operating system they
will work with. A Windows NT machine will not be able to read a disk
formatted for Unix, and vice versa.
Whenever you upgrade your disks, or modify your storage structure in
any way, make sure that you perform your upgrade checklist.

T his chapter covered essential details for using hard disks as storage
devices. First, we talked about hard disk structure, differentiating between
physical and logical disks.
We then discussed IDE hard disk drives. While IDE may not be as fast as
SCSI, it’s very commonly used because of its lower cost. IDE devices are eas-
ier to configure, but you are limited to two IDE devices per IDE controller.
When using two devices on one controller, you need to set one device as
master and the other as slave.
SCSI is the most popular hard disk technology used in servers. It’s fast,
and allows for large numbers of hard disks per machine. SCSI has many dif-
ferent standards, but all are based on backward compatibility. SCSI devices
are somewhat harder to configure than IDE devices, are more expensive, and
require termination. The benefits of SCSI over IDE include greater perfor-
mance, flexibility, and support for internal and external devices. Last, we
looked at some administration tips for managing your disk storage solutions.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Exam Essentials 131

Exam Essentials
Know the difference between a physical and logical disk. The actual
device located within a server, or connected to the server externally, is the
physical disk. A logical disk is defined by its drive letter.
Know what a mapped disk is. A mapped disk is a path on a client
workstation pointing to a network disk or network share. In Windows
Explorer it appears to the client as a physical disk although it is actually
a path to a logical disk or share on another computer.
Know the three configurations for IDE hard disks. Using jumpers, you
can set IDE devices in one of three configurations: master, slave, or cable
select. You can choose cable select if the devices are capable of performing
automatic selection of correct master/slave configuration.
Know the major limitations of IDE in a server environment. One of
the first limitations of IDE is its support for only two IDE devices per
channel. It also lacks the transfer speeds often needed within a server.
Know the three SCSI signaling methods. There are three different sig-
naling methods: single-ended (SE), High Voltage Differential (HVD), and
Low Voltage Differential (LVD).
Know the common SCSI standards. The various SCSI standards also
have varying bus widths, transfer rates, connector types, and cable length
Know the key SCSI configurations. Configuring SCSI devices includes
setting the SCSI IDs and LUNs, and ensuring proper termination.
Understand SCSI IDs and LUNs. Each device must have a unique SCSI ID
that uniquely identifies each component on the SCSI chain. For a SCSI device
performing multiple functions, the LUN is used to identify each one.
Understand SCSI termination. SCSI termination must occur at both ends
of the chain. Usually the controller is terminated at the last device in the
chain. The four basic types of termination are passive, active, FPT, and LVD.
Know the various SCSI cables. Depending on the device, SCSI cabling
can be internal or external. Internal cables follow two forms: standard
ribbon and twisted-pair.
Know the benefits of SCSI over IDE. SCSI provides the following
benefits: faster transfer speeds, support for more devices, and decreased
processor load.

Copyright ©2002 SYBEX, Inc., Alameda, CA

132 Chapter 4  Storage Devices

Key Terms
Before you take the exam, be certain you are familiar with the follow-
ing terms:

Centronics 50 physical disk

FPT (Force Perfect Termination) SCSI (Small Computer
Systems Interface)
High Voltage Differential (HVD) SCSI-1
High Voltage Differential (LVD) SCSI-2
IDE (Integrated Drive Electronics) single-ended (SE)
logical disk termination
Low Voltage Differential (LVD) Ultra 160
LUN (logical unit number) Ultra 320
mapped paths Ultra-3 SCSI
partitioned Wide Ultra-2 SCSI

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 133

Review Questions
1. You are the network administrator for your company. One day, your
boss walks in beaming about a new technology she read about called
IDE. However, your boss forgets what it stands for. What does IDE
stand for?
A. Industrial Drive Electronics

B. Integrated Device Electronics

C. Integrated Drive Electronics

D. Internal Drive Electronics

2. You have three servers on your network. Because the company

wanted to save money, they decided to implement IDE hard disks in
all servers. Each server motherboard has two IDE connectors. What is
the maximum number of hard disks that each server will support?
A. One

B. Two

C. Three

D. Four

3. The current IDE device specifications are based on what industry

A. Attachment Interface

B. American Transistor Association

C. Analogue Terminal Attachment

D. A-Terminal Attachment

4. You are configuring a new server for your company. One of the junior
employees wants to know what types of devices are going to be plugged
into the IDE controllers on the motherboard. Which of the following
is the least likely device you will plug into an IDE controller?

Copyright ©2002 SYBEX, Inc., Alameda, CA

134 Chapter 4  Storage Devices

A. Printer

B. Hard disk



5. You have just installed brand new Ultra ATA/100 hard disks in your
server. However, when you monitor the disk performance, you notice
that you are only getting approximately 25MBps throughput. What is
the most likely problem?
A. The hard disk is defective and must be replaced.

B. The system BIOS needs to be properly configured to recognize the

new hard disk.
C. The motherboard only supports Ultra ATA/33.

D. Nothing. The maximum throughput for an Ultra ATA/100 device

is approximately 25MBps.

6. One of your junior network administrators is trying to configure a

new IDE hard disk. This person does not have much hardware expe-
rience, and is frantically searching for some sort of configuration
method. What do you tell him to use to configure the hard disk?
A. Pins

B. Jumpers

C. Software

D. DIP switches

7. You have just installed a second hard disk into your server. You are
using IDE devices. However, when you boot the machine, it hangs up
during the hard disk detection portion of the POST. You quickly place
the disk into another machine, and it boots fine. What is the first thing
you should check to get the hard disk working properly in your server?

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 135

A. Check the system BIOS to make sure that Ultra DMA parity
checking is enabled.
B. Check the system BIOS to ensure that LBA is enabled for the
hard disk.
C. Check the back of the hard disk to make sure it is properly
D. Check the back of the hard disk to ensure that you have the proper
master/slave configuration.

8. You are installing an old SCSI-1 adapter in a test machine to demon-

strate the original technology to your boss. How many devices will
that SCSI-1 chain support (including the adapter)?
A. Five

B. Six

C. Seven

D. Eight

9. What three types of SCSI signaling have been officially standardized?





10. Your SCSI chain is composed of devices that all use LVD signaling.
You have an older SCSI device that you want to add to the chain,
but you are not sure what type of signaling it uses. What should you
do and why?
A. Do not install it because if it’s an SE device it could ruin the system.

B. Do not install it because if it’s an HVD device it could ruin

the system.
C. A and B are both correct.

D. Go ahead and install it because mixing signaling types does

not matter.

Copyright ©2002 SYBEX, Inc., Alameda, CA

136 Chapter 4  Storage Devices

11. You are the junior network administrator at your company. The
senior administrator is in a rush to fix the server, and tells you to get
a SCSI connector out of the parts box. When you ask what kind, she
tells you to get a connector for an SE device. What type of connector
do you grab?
A. 50-pin narrow

B. DB-35

C. Centronics 68-pin
D. DB-9

12. You are installing a new SCSI hard disk into your server to act as the
boot disk. The only SCSI ID you have available is 5. You configure
the disk for that ID and boot the server. However, the system is boot-
ing from the old disk, not the new one. What could be the problem?
A. The new disk is malfunctioning.

B. The SCSI adapter will only boot from a disk with a SCSI ID of 0.

C. The SCSI adapter will only boot from a disk with a SCSI ID of 7.

D. The SCSI adapter is malfunctioning.

13. What connectors does SCSI-1 use (pick two)?

A. Centronics 50

B. Centronics 68

C. DB-25

D. High-density 68-pin

14. How many devices can be supported by Wide Ultra-2 SCSI?

A. 8
B. 16

C. 12

D. 4

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 137

15. You are in the process of upgrading your SCSI controller in your
server to SCSI-3. You tell your boss that it’s because of the enhanced
bus speed of the newer technology. What bus speeds does SCSI-3
operate at?
A. 10–20MBps

B. 20–30MBps

C. 20–40MBps

D. 30–40MBps

16. You install a new SCSI hard disk into your server. After rebooting the
machine, the new hard disk is not detected. You place the disk in
another machine that has no other SCSI devices, and it responds prop-
erly. What could be the problem?
A. The disk is malfunctioning.

B. The SCSI ID that the disk is using is already being used by another
device in the first machine.
C. The SCSI ID on the disk is set to an invalid number.

D. The LUN ID that the disk is using is already being used by another
device in the other machine.

17. You are using narrow SCSI technology in your server. You already
have maxed out the number of possible devices in your SCSI chain.
You need to expand the server by adding an additional three hard
disks. What do you do?
A. There is nothing you can do, as computers can only have one SCSI
adapter each.
B. Add the hard disks to your IDE connector on the motherboard.
C. Add the devices to the existing SCSI adapter, and place the termi-
nator at the end of the new chain.
D. Add an additional SCSI adapter to your server, and attach the
disks to it.

Copyright ©2002 SYBEX, Inc., Alameda, CA

138 Chapter 4  Storage Devices

18. You are in the process of adding SCSI devices to a new SCSI adapter.
When you attempt to set the first device, you are not sure which ID to
use. You have not changed the default configuration of the SCSI adapter.
What is the most likely SCSI ID assigned to the SCSI adapter cards?
A. 0

B. 9

C. 7

D. 3

19. Your boss has instructed you to create a fast new server for your net-
work. Cost is not an issue, but speed is, and you need to ensure nearly
100 percent uptime. The boss would prefer to be able to expand the
storage capabilities and replace failed disks without taking the server
down, if possible. What type of solution should you implement?
A. Use internal IDE hard disks for the new server.

B. Use internal SCSI hard disks for the new server.

C. Use an external third-party storage solution with hot-swappable

D. The solution your boss wants is impossible to obtain.

20. Which of the following devices is the least likely to be found connected
to a SCSI controller?
A. Hard disk

B. Modem

D. Scanner

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Review Questions 139

Answers to Review Questions

1. C. IDE stands for Integrated Drive Electronics (and sometimes Intel-
ligent Drive Electronics), and was developed as a hard disk standard.
2. D. A single IDE controller will support two devices, and each server
has two controllers, so each server will support four devices.
3. A. IDE devices are based on the ATA (Attachment Interface) standard.
4. A. Of the devices listed, the printer is least likely to use an IDE port.
All of the other devices are commonly connected to an IDE controller.
5. C. Ultra ATA/100 is a fairly new standard, and not all motherboards
support it. In a case like this, the motherboard’s slow link will slow
down the system.
6. B. IDE devices are configured using jumpers. The jumpers are used
to configure the master/slave relationship between IDE devices.
7. D. When using multiple IDE devices, ensure that you have the proper
master/slave relationship or the disks will not function properly.
8. D. A SCSI-1 chain will support a maximum of eight devices. Each
must have a unique SCSI ID number, from 0 to 7.
9. A. There are three different types of SCSI signaling: single-ended (SE),
high voltage differential (HVD), and low voltage differential (LVD).
10. B. You can mix SE and LVD devices on the same chain, although you
need a terminator that can handle both signaling methods. Never mix
HVD with the other two signaling methods.
11. A. SE signaling uses a 50-pin narrow connector.

12. B. Some SCSI adapters will only boot to a disk if it is set to ID 0. In

the case above, reconfigure the SCSI IDs and you should be okay.
13. A, C. SCSI-1 uses a Centronics 50 or a DB-25 connector.

14. B. Wide Ultra-2 SCSI supports a maximum of 16 devices.

15. C. SCSI-3 operates at bus speeds of 20–40MBps.

16. B. Each SCSI device on a chain must have a unique SCSI ID.

Copyright ©2002 SYBEX, Inc., Alameda, CA

140 Chapter 4  Storage Devices

17. D. Computers can have more than one SCSI adapter. If the first one
is full, add a second to accommodate the additional devices.
18. C. Generally speaking, SCSI adapter cards are by default assigned
SCSI ID 7. This makes the card the highest priority item in the SCSI
19. C. Internal SCSI and IDE disks are not hot-swappable. Although
third-party storage solutions can be expensive, they are also very fast,
and often have built-in fault tolerance and redundancy as well.
20. B. A modem is not considered a SCSI device.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Fault Tolerance
and Redundancy

 2.2 Configure RAID

 7.1 Plan for disaster recovery

 Plan for redundancy (e.g., hard disks, power supplies, fans,
NICs, processors, UPS)
 Use the technique of hot swap, warm swap, and hot spare
to ensure availability
 Use the concepts of fault tolerance/fault recovery to create
a disaster recovery plan
 Develop disaster recovery plan
 Identify types of backup hardware
 Identify types of backup and restoration schemes
 Confirm and use off site storage for backup
 Document and test disaster recovery plan regularly, and
update as needed

Copyright ©2002 SYBEX, Inc., Alameda, CA

I t is amazing how many businesses rely heavily on their server
yet invest so little in making sure that the server is highly available. High
availability and redundancy go hand in hand. In its essence, redundancy
is having multiples of the same device present in a computer so that if one
device fails there will be another to take over. The obvious benefit of pro-
viding redundancy is to ensure that your server stays up, which translates
into your company being able to go about its daily business.
This concept of redundancy has expanded over the last several years to
include a wide array of components and services. By including redundancy
within your server configuration, you are increasing the availability of the
server. When a component within the server fails, the redundant component
will take over and provide a seamless transition. Because the server is nor-
mally the heart of a network, great care should be taken to ensure that it is
highly available.
Throughout this chapter we will discuss the various ways that you can
prevent system failure by practicing common redundancy techniques and by
taking advantage of clustering and RAID technology. We will also spend
some time evaluating the various levels and forms of RAID.

Managing Fault Tolerance

W hen an error occurs within a computer, it can be a result of numer-
ous variables. You will remember from the earlier chapters that a multitude
of components and processes are working within the computer case. When
resources are low or shared, computer systems can become unstable and
even stop responding. This often results in system failure and the need for a

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 143

reboot. However, this is not the only type of problem that can occur within
a computer. These are the three broad fault categories you might see:
Computer Hardware Faults occur when a hardware component fails.
For example, a network card fails, resulting in no access to the server via
network communication. Hardware fault tolerance can provide redun-
dancy by supplying several (or at least two) network cards. The network
cards can be configured to monitor each other. When the primary net-
work card fails, the secondary card can take over.
Software Faults can also bring a server to a halt. By providing mecha-
nisms to support operation despite possible software errors and/or
failures, you increase your availability. These mechanisms can include
monitoring tools to assess system resource utilization or redundant
programs to ensure data access and manipulation.
System Level Faults occur in areas that are not computer based, such as
sensors, lights, diodes, etc. These components, although they may not be
as critical as computer hardware and software, still play an important
role in system operation. Another example of system-wide fault tolerance
would involve monitoring other network components, such as switches
or routers.
Maintaining fault tolerance is maintaining the ability to accept a fault
within one component of a subsystem without losing services to another.
That said, the primary objective in fault tolerance is eliminating any single
point of failure (SPOF). Depending on the server function, SPOFs can vary.
For example, a high-availability web server will have multiple possible Inter-
net connections, so that if the primary Internet connection fails, then a sec-
ondary and or tertiary connection can take over. In contrast, a print server
will not commonly have a web connection but could have several printer
connections and/or multiple links to remote printers.
Depending on the demands of the server at hand, redundancy offers sev-
eral possibilities. At the planning stage, it must first be decided at what level
the server needs to be available. Hot site servers, such as those in hospital and
police networks, must be available always, regardless of any possible disaster
or problem. Hot servers are the most expensive and fault tolerant. Warm
servers are designed to be fault tolerant most of the time. They contain sev-
eral redundant components, usually in what is deemed the likeliest areas of
possible faults. Warm servers cost more than cold servers but not nearly as
much as hot servers. These servers will be fault tolerant most of the time but

Copyright ©2002 SYBEX, Inc., Alameda, CA

144 Chapter 5  Fault Tolerance and Redundancy

still can go down from time to time. Cold servers contain few if any redun-
dant components. Cold servers can and often do fail; pricewise, they are the
most affordable. What you want to do is achieve a balance between cost and
reliability. Ask yourself, “What is the use of the server? What key compo-
nents are in use and stand a chance of failure?” These are the components
that should be part of your fault tolerance plan.
Eliminating every possible SPOF and having maximum availability would
be excellent—but expensive and nearly impossible to implement. If you
think about all the possible SPOFs within a single server, this would mean
installing multiples of each key component. Then, to ensure that there would
be full system protection, you would have to use a backup server (you will
learn more about this clustering of servers later in this chapter).
Consideration must also be given to electrical requirements. An uninter-
ruptible power supply (UPS) is a must, but what if the power fails for more
then the expected battery life of the UPS? Many high-availability systems
employ an electrical generator to provide power in case of a lengthy outage.
Now you might want to provide a backup generator for the first generator.
As you can see, this becomes expensive and can get carried away very quickly.
Instead, most companies will implement a warm server.
In the real world, where the value of a dollar is taken into consideration,
redundancy focuses on the commonly used components as well as those that
might be susceptible to failure. These common components, as well as the
reason for their redundancy, are explained in the next section.

Common Redundant Components

Within a server are several components that, due to their importance or
stress load, should have redundancy. By providing redundancy to these
components, you can improve availability. We will take a closer look at
them within this section.

Network Cards
Network cards within a server play an integral role in the network. Without
a NIC, your server is still operational but it suddenly becomes a stand-alone
computer. All the files and resources that are shared become inaccessible to
client computers. Current network cards are, for the most part, inexpensive.
Many servers will contain two network cards; most hold several. The bene-
fits of more than one card are numerous. In terms of redundancy, more than
one connection from the server to the switch or hub provides redundant

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 145

paths. If one network card were to fail, then a secondary or tertiary card
would be available to take over the requests. This is referred to as adapter
fault tolerance. Configurations can include all network cards working
together as a team (adapter teaming) and handling requests as one, even
though there are several cards working together.
The benefits to this are obvious. First of all, should the primary card fail,
the second card can take over without any intervention. Adapter teaming
can also provide a certain level of load balancing as network requests can be
distributed evenly between the cards (in turn eliminating the possibility of a
single network card becoming a bottleneck).

Power Supplies
Redundant power supplies are becoming increasingly common within
servers. It is fairly common to have a power supply fail. Remember, the con-
version from AC (alternating current) to DC (direct current) occurs within
the computer’s power supply. Component failure and/or fan failure is a con-
cern. Servers containing multiple power supplies are configured to monitor
the primary power supply; if needed, the secondary power supply will take
over from a failing or failed primary supply. However, if the problem is poor
AC power entering the server, then redundant power supplies will not solve
your problem.

Hard Disks
Hard disks are one of the most common components to be seen in a redun-
dant configuration and one of the most common components within a server
to fail. If you can afford to provide only one area of redundancy, then it
should be the hard disks. Remember, your hard disks take the brunt of the
daily stress in a server environment. The hard disks also contain all of the
information that is stored on your network. If any other component within
the server fails, it can be changed out with nothing more than some lost pro-
ductivity time. If the hard disk fails, then all the data and information that
was stored on the disk is also lost. This is why backups are so important.
Backups allow you to restore in the event of data loss. But think of the time
it may take to perform a restore. This is lost productivity time. By imple-
menting hard disk redundancy, you can greatly reduce the time it takes to
bring a server back up in the event of hard disk failure—in some instances,
no downtime at all will be experienced. Hard disk redundancy raises a whole
new concept called RAID. We will be looking at RAID later in this chapter.

Copyright ©2002 SYBEX, Inc., Alameda, CA

146 Chapter 5  Fault Tolerance and Redundancy

Cooling Fans
It goes without saying that cooling is critical within a computer. A server
will generate even more heat than a standard desktop, so cooling is a more
serious consideration in servers. Running several high-spin-rate hard disks,
multiple adapter cards, high-performance processors, and multiple power
supplies generates large quantities of heat. Cooling fans are available in a
variety of forms today. Dynamic fans that are dedicated to a specific com-
puter component, such as a hard disk, are available at minimal cost and are
well worth the investment.
Many server environments contain numerous fans, as well as groups of
fans working together. This form of redundancy ensures that when a fan fails
(and being a mechanical component, it will fail at some point), the failure
will not result in rapid overheating and eventual system instability. Most fan
speeds (RPMs) are also controlled. If needed, the fans can be speeded up or
slowed down to control the temperature.

Internet Connection
Where would we be without the Internet? Most businesses rely on the
Internet for their daily operations more than they realize. If daily operations
are reliant on the Internet, then precautions should be taken to ensure con-
nectivity. Precautions include providing multiple connections and different
forms of connectivity. This means that if the T1 connection fails due to issues
with the provider, then Internet access can be gained through another pro-
vider, another service (such as DSL), or—if all else fails—through the good
old dial-up modem. As much as we complain about slow modems, it is better
to have dial-up than nothing.

Clustering Technology
Reliability of network servers has become critical to the success of many
businesses. Most businesses today have resources, applications, and services
hosted on network servers that are crucial for their day-to-day operations.
This means that these resources need a high level of availability. One of the
key technologies available to meet this requirement is clustering. Clustering
servers so they operate as a single server can increase the availability of
resources, applications, and services to an impressive 99.999 percent. Not
only does it provide an economical solution for fault tolerance in the event
of server failure, but it also makes planned outages more convenient.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 147

What Is a Cluster?
A cluster is a group of computers that work together as one and logically
appear to be a single system to users on the network (see Figure 5.1). It is a
combination of both hardware and software solutions. Clustering allows
you to link two or more systems together so that if one should fail the other
is ready to automatically assume its workload. In the event of server failure,
applications, services, and resources are migrated to a remaining cluster
member by the cluster software and are restarted.

FIGURE 5.1 Clustering

Public network Public network

connection connection

Server 1 Server 2

Shared disk

Private network connection

Clustering can be a viable solution for mission critical servers, whether

they are transaction servers, database servers, or mail servers. The cost of
implementing a cluster configuration can be far outweighed by the cost of
lost data and server downtime. The benefits of such a configuration include
fault tolerance and increased availability, load balancing, and scalability.
The benefits you achieve by implementing a cluster will of course depend on
how you configure it (for example, active/active or active/passive).
Fault Tolerance and Increased Availability In a cluster configuration,
at least one system is on standby should the primary fail. When the
primary system fails, or a component on that system fails, the cluster
software responds by moving the resources from the failed system to
the standby. Downtime can be decreased to just a few seconds, thereby
providing fault tolerance and increased availability.

Copyright ©2002 SYBEX, Inc., Alameda, CA

148 Chapter 5  Fault Tolerance and Redundancy

Load Balancing By implementing a clustering solution, multiple servers

can run the same services. When clients make a request for the service, the
requests can be balanced among the different servers. The servers appear
to the clients as a single logical server and the clients are unaware of which
physical server actually responds to the request.
Scalability Depending on the clustering software you are using
and the operating system you are running it on, you may have some
level of scalability with your cluster. For example, if you are running
Microsoft Windows 2000 Datacenter Server and Microsoft Cluster
Service, you can add up to four servers to a single cluster. Scalability
can also be achieved within a cluster by adding more resources and
components to a system.

The availability of a cluster and its resources is dependent on the configuration

of the network environment. So, when setting up a cluster, you also need to
identify any single points of failure that might impact the resources’ availability.

You can configure an active/active cluster or an active/passive cluster.

The model that you choose will have an impact on how the cluster operates.
In an active/active configuration, each of the servers has its own workload
but is also ready to assume the workload of another cluster member in case
of failure. For example, you might have a database application running on
one cluster member and an e-mail application running on another member.
Should either one fail, the other is ready to assume the failed server’s work-
load while still maintaining its current load. In an active/passive configura-
tion, one server assumes the workload while another server sits idle, ready to
assume the workload should the active server fail.
Clustered servers have to have access to the same data. Looking at the
previous examples, if the database application fails over to the second
cluster member, that member will need access to the actual data to continue
servicing client requests. This is achieved using shared storage. The cluster
members will all be connected to a shared storage device that they all have
access to. In most cases, this will be an external disk or multiple external
disks. The cluster software is responsible for determining which cluster
member has access to the data on the shared disk.

The servers in a cluster provide fault tolerance through failover and failback.
In the event of server failure, component failure, or service failure, the

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 149

workload of the failed cluster member will automatically be assumed by

another server. The applications and services running on the failed member
will be restarted on another cluster member, which will assume the respon-
sibility of servicing client requests; this is known as a failover. The cluster
members use heartbeat messages to monitor one another’s online status and
the status of clustered resources. Once the failed member is back online, the
workload can be configured to be automatically transferred back to the now
operational server, a process known as a failback. The only downtime that
will be experienced is the amount of time it takes for the workload to be
automatically transferred between cluster members. The end user sees the
cluster as a single system, and in the event of failure, user requests are trans-
parently redirected to another server.

What Is RAID?
A server’s disk subsystem is one of the most common components to fail.
With this in mind, you will want to implement some form of disk subsystem
fault tolerance when building a reliable network server. RAID (Redundant
Array of Independent [or Inexpensive] Disks) is a group of hard disks that
collectively acts as one storage system, providing tolerance to failure of a
disk within the array.
There are many benefits to using RAID within a server. Providing a means
for high availability within the server is always the primary objective. RAID
allows for data to be highly available regardless of disk failure. In a complex
disk array, one hard disk or more could fail and the server will still run seam-
lessly to the end users. This ability of combining multiple hard disks into a
fault tolerant array is what makes RAID so appealing to server technicians.
Another benefit is speed. Having data stored on multiple disks allows for
the disks to all write information at one time. This speeds the writing process
considerably over a single-disk system.
The main disadvantage of RAID is the cost of implementation. The cost
will vary between implementations depending on factors such as the number
of disks required, the amount of disk space required, and the level of RAID
you choose to implement. Keep in mind as well that RAID does add a level
of fault tolerance to your network server disk subsystem and data but does
not provide a 100 percent fault tolerant solution because most levels of
RAID can recover from failure of only a single disk.
There are two forms of RAID: hardware based and software based.
Hardware based RAID uses a controller card (similar to the SCSI card) that

Copyright ©2002 SYBEX, Inc., Alameda, CA

150 Chapter 5  Fault Tolerance and Redundancy

connects to the motherboard through an expansion slot, or can also be an

integrated component of the motherboard. In hardware based RAID the
controller card contains a BIOS chip that maintains the disk information.
Hardware based RAID is often preferred to software based RAID because
the responsibility of controlling the disks is taken on by the controller card
rather than the operating system and processor. This frees valuable resources
within the server. Software based RAID relies on the network operating
system to configure and control the information being sent and accessed on
the disks.
Later in this section, we will discuss both forms of RAID in great detail,
but before we look at the levels of RAID, an understanding of RAID terms
must first be established.
Array An array is a grouping of disks contained within a single imple-
mentation of RAID. Different levels of RAID will require a different
number of disks within an array. For example, implementing RAID level 1
requires 2 disks within the array while RAID 5 requires a minimum of 3
with a maximum of 32.
Striping Striping is a method of spreading data across several disks.
There are tremendous performance gains with striping because data is
written and read from the disks simultaneously. The performance of strip-
ing is measured by the size of the stripe. Different levels of RAID offer
different sizes of stripes, such as bits, bytes, and blocks. The larger the
data stripe, the better the performance of the stripe.
Mirroring Mirroring is a simple concept: Data is duplicated on two sep-
arate hard disks. Should one of the disks fail, then the second disk will
take over seamlessly. The controller card does this changeover between
disks should a failure arise. Write performance is not very good in a mir-
rored array because all data has to be written twice (once for each disk).
Duplexing Duplexing is similar in concept to mirroring. Data is dupli-
cated on two separate disks but duplexing adds an extra level of fault
tolerance through the use of separate controller cards.
Parity Parity is added to certain levels of RAID. The parity bit is used as
a form of error checking for the array. Each data byte written to the array
is given a parity bit. The controller card uses the parity bit to validate the
integrity of the information being written to the disks. If the data has been
compromised, the controller card will replace the data automatically with
information from the other disks.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 151

Volume In terms of RAID, the volume is the total amount of logical disk
space within the array. For example, if you were to implement RAID 5
using four physical disks, combining 3GB of free space from each one, the
RAID volume would be 12GB (four disks × 3GB). This means that there
is 12GB of free storage space within the array. Keep in mind that this
calculation does not take into account the space needed for the parity

Levels of RAID
There are numerous levels of RAID that can be implemented in a hardware
based RAID array. Each level has its own benefits and drawbacks. The most
common levels are RAID 1 (disk mirroring) and RAID 5 (stripe set with
parity). The Server + exam will test your knowledge on seven of these levels.
First let’s look at the levels in detail, and then we can compare them in a
chart to see the benefits and drawbacks.
Level 0 RAID 0 provides no fault tolerance at all. Data is split across
hard disks, resulting in fast data throughput but no safety. If a disk were to
fail, then the data would become inaccessible. RAID 0 is often referred
to as striping. Level 0 is one of the implementations of software level
RAID, and will be discussed later in the chapter.
Level 1 RAID 1 is called mirroring. In mirroring, two disks are used
and data is copied (mirrored) from one disk onto another. When one disk
fails, there is an identical second hard disk to take over. Most ordinary
servers will use RAID 1 for fault tolerance. Level 1 is a common imple-
mentation in servers and can be used with IDE disks as well as SCSI
hard disks.
Level 3 RAID level 3 is striping bits of data across several disks with
parity information stored on one disk. A major concern with this array
is that the parity disk is a SPOF. Should the parity disk fail, then the entire
array will halt. RAID level 3 requires at least three hard disks (including
the parity disk). There is also an increase in workload placed on the parity
disk because each time a write operation is performed this disk is accessed.
Level 4 RAID 4 stripes data as bytes across several disks with parity
information stored on one disk. Parity data is updated on each write

Copyright ©2002 SYBEX, Inc., Alameda, CA

152 Chapter 5  Fault Tolerance and Redundancy

request, which can hamper performance. The same SPOF issue seen in
level 3 is also a concern in level 4. RAID level 4 also requires at least three
hard disks. The benefit over level 3 is that the data being written to the
disks is in larger units (bytes over bits).
Level 5 RAID level 5 is commonly referred to as striping with distributed
parity. Similar to levels 3 and 4, RAID 5 offers a more advanced parity,
which is striped across multiple disks. This ensures that if the parity disk
were to fail, the array would not fail. On the negative side, because parity
data must also be written to each disk in the array, performance is slower.
Level 0+1 RAID 0+1 (sometimes referred to as RAID 10) is a dual array
that takes the best of level 0 and level 1. Multiple mirror sets are used,
which are then configured in a striped set (requiring a minimum of four
disks). RAID 0+1 offers high data-transfer speed with data protection.
Level 0+5 This level of RAID is composed of multiple RAID 5 sets
connected in a single array. The benefit of this complex structure is that
multiple disks could fail across several sets and still the entire array
would stay active. The cost of such a structure would be mind-boggling.
Now that we have a better idea of each of the main RAID levels, we can
look closely at the benefits and drawbacks of each level and compare them
(see Table 5.1). It will be important for you to know these and be able to
make distinctions between levels, not only for the purposes of the exam but
also for day-to-day hands-on activity.

TABLE 5.1 Levels of RAID

Level Benefits Drawbacks

RAID 0 Fast No data protection

RAID 1 Fast and easy to set up Controller a SPOF and

is slower than 0

RAID 3 Uses parity Parity disk a SPOF and

stripes small data
units (bits)

RAID 4 Uses parity Parity disk a SPOF

and stripes larger
data units

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 153

TABLE 5.1 Levels of RAID (continued)

Level Benefits Drawbacks

RAID 5 Best cost/performance Only supports SCSI

disks in hardware level

RAID 0+1 High transfer speeds Expensive and often

difficult to configure

RAID 0+5 Best data protection Extremely expensive

RAID Disks
In most applications SCSI disks are the disks of choice in a RAID config-
uration, but IDE disks can also be used. If you remember from Chapter 4,
“Storage Devices,” IDE supports only two disks per controller, and therefore
IDE is not often used within a RAID configuration. However, with software
RAID, mirroring can be implemented on one or more IDE disks. This will
allow for data protection. Normally SCSI disks are used. Many manufactur-
ers sell combination RAID and SCSI controller cards. These amazing (and
normally very costly) multi-channel devices allow you to select whether the
channels will function as RAID or SCSI controllers. Many mid-range to
high-end servers utilize these controllers.

RAID Controllers
RAID disks are just one component to consider when implementing RAID.
You will also need to consider the RAID controller if you are implementing
hardware level RAID. RAID controllers perform functions such as calculat-
ing the parity information and caching of information. RAID controllers
come with their own processor to perform parity calculations, which means
the RAID system will no longer be dependent on the CPU in the server.
In terms of caching, the controller can cache information that will cache data
that is waiting to be read from or written to the disks in the array. When
choosing a RAID controller, you need to consider the type of disk subsystem
(SCSI or IDE), the type of RAID you plan to implement (not all RAID con-
trollers support every level of RAID), and the size of the on-board cache. The
size of the on-board cache will be determined by the type of data being stored
on the array and the expected workload on the controller.

Copyright ©2002 SYBEX, Inc., Alameda, CA

154 Chapter 5  Fault Tolerance and Redundancy

RAID Cache
Depending on the workload, a RAID controller could become a bottleneck
trying to perform all of the read/write operations. Most RAID controllers
come with on-board cache to eliminate this possibility. If the RAID control-
ler receives a request that it cannot immediately perform, the request can
be temporarily placed in the cache.

Software RAID
Some operating systems allow you to configure RAID without the need for
special hardware. The operating system will come with some utility that
will allow you to configure software level RAID, which is known as software
RAID. For example, using Disk Management, a tool included with Win-
dows 2000, you can implement software RAID through the operating
system. The previous section gave you a brief introduction to the different
levels of RAID that can be implemented. The following section will describe
in more detail the features, benefits, and drawbacks of software level RAID.
Throughout the section we will pause and examine the advantages and the
disadvantages of most of the common software RAID levels. You will find
these highlighted evaluations in shaded sidebars within this section.
RAID 0, also known as disk striping, can be implemented on a server but
it does not offer any fault tolerance. So if your servers are hosting mission
critical data, this will not be an appropriate solution.
With RAID 0, data is broken down into blocks and written across multi-
ple hard disks, which increases performance. Performance is also increased
because there is no parity overhead. However, should a disk within the array
fail, all data is lost and is only recoverable by restoring from a backup copy.

RAID 0 Advantages

 Improved performance—because data is spread across multiple disks, I/

O performance in increased.

 Minimal hardware—RAID 0 can be implemented on two disks.

 Easy implementation—it is easy to implement and can be done using a

tool that comes with the operating system.

 Zero disk overhead—because there is no parity information, there is no

disk overhead associated with RAID 0.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 155

RAID 0 Disadvantages

 No fault tolerance—if a single disk fails, all data is lost and must be
restored from backup.

 Increased single point of failure—most implementations of RAID 0 can

support up to 32 disks. However, the more disks you have, the more
points of failure. If any of the disks within the array fail, all data will
be lost.

Because RAID 0 offers no fault tolerance, it should never be used for data that
is mission critical.

RAID level 1 is also known as disk mirroring. This is one of the most
common implementations of RAID in a server environment. With disk
mirroring, two disks are required so that data from one disk can be copied
or mirrored onto a second disk. Each time a write is made, it is duplicated to
the second disk in the mirrored set. If the first disk fails, the data can be
accessed from the second disk in the mirrored set (see Figure 5.2).


Disk 1 Disk 2

C: C:'

Disk Controller

RAID level 1 has little impact in terms of performance. You will not
see any increase in performance when reading from the disk and you may see
a decrease in performance for disk writes because the data now has to be
written to two different disks.

Copyright ©2002 SYBEX, Inc., Alameda, CA

156 Chapter 5  Fault Tolerance and Redundancy

In terms of fault tolerance, RAID level 1 can withstand the failure of one
disk without data loss. This implementation however cannot withstand
the loss of a second disk, so when disk failure occurs, it is important to
replace the disk as quickly as possible.
A variation of RAID 1 is disk duplexing. It is similar to disk mirroring but
provides an additional level of fault tolerance. With disk mirroring, the disks
in the array use the same disk controller. Should the disk controller fail, both
disks fail as well. With disk duplexing, each of the hard disks has a separate
controller, adding yet another level of fault tolerance.

RAID 1 Advantages

 Easy to implement fault tolerant options. Since data is duplicated on

two disks, it is readily available should one of the disks fail.

 Disk duplexing—this process can be implemented to eliminate the disk

controller as a single point of failure.

 Fault tolerance for system partition—normally, boot files cannot be

written across multiple disks. RAID 1 allows you to implement fault
tolerance for the system partition.

 It offers a relatively low-cost fault tolerant solution because only two

disks are required.

RAID 1 Disadvantage

 Disk overhead costs—the major disadvantage of RAID 1 is the disk

overhead or the cost per megabyte. A large amount of disk space can be
consumed mirroring one disk onto another. RAID 1 has the highest disk
overhead of any RAID implementation.

Some implementations of RAID allow you to add a hot spare, which can be
used if one of the disks in the array fails.

The second most common level of RAID is RAID level 5, also known as
striping with parity. It requires a minimum of 3 disks and most implementations

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 157

support a maximum of 32 disks. Data that is written to the array is stored

across multiple disks. This level of RAID provides fault tolerance by adding
parity information that can be used to regenerate the data should a disk in
the array fail. The volume size for the array is determined by the smallest disk
(all disks in the array must be the same size). For example, if you have three
disks, 10GB, 20GB, and 40GB in your array, 10GB from each disk will be
used because this is the smallest disk size. The total volume for the array
would be 30GB, with 20GB available for data storage because 10GB is
reserved for parity.
In terms of fault tolerance, RAID level 5 can withstand the loss of a single
disk and use the parity information to recover the lost data. If more than
one disk in the array fails, all data is lost.
Not only does RAID level 5 provide fault tolerance for data, but it can
also increase performance because it is faster to read data from multiple
disks than it is from a single disk. You may see a slight decrease in write per-
formance because the parity information has to be calculated. The trade-off
for fault tolerance is a slight increase in server overhead. A portion of your
disk volume will be used to store the parity information. To calculate the
amount of disk space for the parity information use the formula 1/x where x
is the number of disks in the array. For example, if you have four 20GB disks
in your RAID implementation, the total volume is 80GB but 20GB is
reserved for parity information (1/4 of the total volume space).

Level 5 Advantages

 Fault tolerance—obviously, one of the main advantages to RAID 5 is

fault tolerance. It can withstand single disk failure while still providing
access to the data, using the parity information to re-create the data
stored on the failed disk.

 Increased performance—RAID 5 offers an increase in performance

because data, including the parity information, is written across
multiple disks.

 Replace without shutting down—depending on your server configura-

tion, if it supports hot swapping, you will be able to replace the failed
disk without shutting down the server.

 Low disk overhead—the disk overhead for RAID 5 is considerably lower

than that of RAID 1.

Copyright ©2002 SYBEX, Inc., Alameda, CA

158 Chapter 5  Fault Tolerance and Redundancy

RAID 5 Disadvantages

 Restrictive array—the system and boot partitions cannot be contained

on a RAID 5 array.

 Limited performance—there is a decrease in write performance because

the parity information has to be written across multiple disks every time
a write operation is performed.

 Disk Overhead—Even though the overhead is less than that of RAID 1,

one disk space is still lost for parity information.

Software Versus Hardware

As you now know, RAID can be implemented through software and there it
is configured and then managed through the operating system or by using
specialized hardware. When determining the level of RAID to implement,
you will need to look at the benefits and drawbacks of each type as well
as your requirements (costs, performance, capacity, and fault tolerance).
Keep the following points in mind when you are deciding what type of
RAID to use.
Costs Software level RAID reduces costs because it does not require any
specialized hardware.
Performance Software level RAID is handled by the server’s processor,
which can decrease performance. Hardware level RAID takes the load off
the server’s processor with the use of a specialized controller card.
Capacity Software level RAID is dependent on the operating system and
consumes server resources. Hardware level RAID is not dependent on the
operating system.
Fault Tolerance Software RAID does have certain limitations, such as
the system and boot partition being unusable in RAID level 5. Hardware
level RAID can offer faster data transfer rates and can support technolo-
gies such as automatic failover of a failed disk to a spare and hot spares.

You can implement different levels of RAID on a single server. For example
you may choose to use RAID 1 for the system and boot partition while imple-
menting RAID 5 for data.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Managing Fault Tolerance 159

Unlike hardware level RAID, software level RAID supports both SCSI and IDE
disks in a single array.

RAID for Bargain Hunters

Not long ago I did some pro bono consulting for a small independent school
in Massachusetts, where the administration had become concerned about
data security. Wisely perhaps, the school’s administrative network was
designed to be completely separate from the student labs, so I was spared
from dealing with dozens of wickedly clever teenage saboteurs. Instead, I
had only to deal with one cost-conscious head of school. The administrative
network was running Windows NT on a server with five physical hard disks.
The head wanted me to set up the server so that if one of the disks failed, the
data would not be lost. Oh yes…and it had to cost nothing.

Hardware Level RAID was clearly out of the question unless I could con-
vince somebody’s grandmother to donate a RAID controller. But I had the
solution. By applying software level RAID through an existing NT utility, I
implemented disk striping with parity (also known as software level
RAID 5). Software RAID relies on the operating system to control disk reads
and writes, so the processor takes a bit of a hit and some disk space is
lost to parity information; luckily, these were not major concerns on this
network. In fact, one of the math instructors reported a slight increase in
performance, probably due to faster reads from the multiple disks. I had
delivered a good level of fault tolerance: If one disk in the five-disk array
failed, data could be recovered through stored parity information.

When I explained all this to the head, she graciously thanked me and
assured me that my work was worth every penny.

Hot Plug
Hot plug is an amazing technology that allows you to add disks to a server
while it is running. This technology calls for some special components but is
well worth the added expense. Servers are built with a backplane and rail sys-
tem that allows for the disks to slide in. Once fully inserted the disks come into

Copyright ©2002 SYBEX, Inc., Alameda, CA

160 Chapter 5  Fault Tolerance and Redundancy

contact with the backplane where connectivity is established. This connectivity

can occur through metal electrical connections or through fiber (light) connec-
tions. The operating system should then identify a new disk has been added.
In the modern server with a hot plug backplane, normally each disk con-
tains lights on the front. Much like a traffic light, green means go and red
means stop. These hot plug disks can be monitored through the lights. When
a disk fails or is having problems, the red light will shine, thus giving you a
quick visual note that you should check things out. If the disk has indeed
failed and needs to be replaced, it can be swapped out by sliding it out and
replacing it with your spare. The RAID controller will sense the new disk
and begin to rebuild it with the data from the other disks. This rebuilding
process involves installing the disk as part of the RAID array and transfer-
ring the data over. The length of time that this process will take varies with
the size of the disk, type of RAID configuration, as well as quantity of data.
To the end users this process will be seamless. They will not be aware that
there has been any issue with a disk replacement.

Hot Spare
In RAID configurations it is often advisable to have a hot spare on hand. Hot
spares are extra hard disks (matching those in use within the server), which
can be installed when needed. By having a hot spare on hand, you can be
assured that if and when a disk fails within the server, a replacement can be
installed and integrated into the RAID configuration as quickly as possible.
Unfortunately the hot spare idea is not an effective use of money. There is
a possibility that you may never use the hot spare disk and it will sit in a
storage cabinet until it becomes obsolete.
With a RAID 1 configuration having a hot spare makes sense. After all, in
this implementation of RAID, you only have two disks to work with. When
one fails, you are down to only one disk—and if you recall, RAID 1 can only
recover from a single disk failure. If you then need to send out the failed disk
for servicing and wait for it to return, you suddenly are spending an extended
period of time with no data protection.

In this chapter we discussed server fault tolerance and the different
options that can be used to increase server availability. Providing redundant
components is one of the simplest ways to avoid server downtime.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Exam Essentials 161

To provide for complete server fault tolerance, not just fault tolerance of
individual components, servers can be set up in a cluster configuration. This
provides a high level of availability for servers as well as the applications and
services they are hosting. In a cluster configuration, two or more servers
operate as one—should one server fail, another is ready to automatically
assume its workload.
The disk subsystem is one of the most common components to cause
server failure. Providing redundancy and fault tolerance for data stored on
a server’s hard disk can be accomplished by implementing some form of
RAID, using the operating system or specialized hardware.

Exam Essentials
Recognize the three general categories of server faults. Faults that can
occur in a server environment include hardware faults, software faults,
and system-level faults.
Common redundant components. There are many ways to provide
server fault tolerance. One way is to implement common redundant
components. Due to their importance in a server environment, you should
consider implementing redundancy for the following components: Net-
work interface cards, power supplies, processors, cooling fans, and
hard disks.
Understand how clustering provides fault tolerance. To provide fault
tolerance for a server and the applications, services, and data it is hosting,
you can implement clustering technology. With clustering, two or more
servers act as one. If one server fails, another server is ready to auto-
matically assume its workload. Clustering provides fault tolerance,
scalability, and load balancing.
Understand RAID. RAID, or redundant array of inexpensive disks, is a
group of hard disks that collectively act as one storage system to provide
fault tolerance for a server’s disk subsystem. RAID can be implemented
through specialized hardware or through the operating system.
Understand the commonly used levels of RAID. The most commonly
used levels of RAID are 1 and 5. RAID level 1, also known as disk mir-
roring, takes data from one disk and mirrors in onto another disk. RAID

Copyright ©2002 SYBEX, Inc., Alameda, CA

162 Chapter 5  Fault Tolerance and Redundancy

level 5, also known as disk striping with parity, writes data across multi-
ple disks and uses parity information to re-create the missing data in the
event of disk failure.
Understand Hot Plug and Hot Spare. With the use of specialized
components, hot plug allows you to add disks to a server while it is
still running. Hot spares are extra hard disks (matching those in use
within the server), which can be installed when needed.

Key Terms
B efore you take the exam, be certain you are familiar with the
following terms:

adapter fault tolerance mirroring

array parity
availability RAID (Redundant Array of
Independent [or Inexpensive] Disks)
cluster RAID 1
computer hardware faults RAID 5
disk controller redundancy
duplexing single point of failure (SPOF)
failback software faults
failover software RAID
fault tolerance striping
heartbeat system level faults
hot plug volume
hot spare

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 163

Review Questions
1. What is the minimum number of disks needed to implement RAID
level 5?
A. 2

B. 3
C. 4

D. 1

2. Your boss is planning on implementing hardware level RAID

because he understands that is more reliable than software RAID.
Which of the following statements is true? (Choose all that apply.)
A. It is not necessarily more reliable but it does provide better
B. Software level RAID is less costly and provides better overall
performance because the RAID subsystem is controlled by the
operating system.
C. Hardware level RAID will be more costly to implement because
specialized hardware is required.
D. Software level RAID will be more costly because a special version
of the operating system needs to be purchased.

3. Of the following, which implementation of RAID combines data

striping across disks with mirroring?

B. RAID 1+5

C. RAID 0+1

4. Of the following, which level/levels of RAID offer no fault tolerance?

Copyright ©2002 SYBEX, Inc., Alameda, CA

164 Chapter 5  Fault Tolerance and Redundancy





5. You create a RAID 5 array that consists of two 10GB disks, a 20GB
disk, and a 40GB disk. What is the total volume space for the array?
A. 0GB
B. 80GB

C. 40GB

D. 50GB

6. You create a RAID 5 array that consists of two 10GB disks, a 20GB
disk, and a 40GB disk. What is the total amount of space available for
storing data?
A. 40GB

B. 60GB

C. 20GB

D. 30GB

7. Which of the following levels of RAID has the lowest disk overhead?



E. RAID 10

8. Which of the following are advantages of software level RAID as

opposed to hardware level RAID?

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 165

A. Both SCSI and IDE disks can be used in the same array.

B. Software level RAID provides improved performance.

C. Software level RAID is less costly because it doesn’t require any

special hardware.
D. Software level RAID is dependent on the operating system,
therefore making it more fault tolerant.

9. Your network server is running a business critical database appli-

cation. You want to minimize server downtime and increase the
availability of the application. Which of the following fault tolerant
options will best meet your needs?
B. Adding an additional processor

C. Clustering

D. Raid 0+1

10. You have configured your network servers in a cluster. Which

components have you provided redundancy for?
A. Network cards

B. Processors

C. Servers

D. Hard disks

E. RAID systems

11. You set up two servers in a cluster configuration. Each server has its
own workload—one is running a database program while the other
is running a mail service. Each server is also ready to assume the
other’s workload in the event of failure. What type of cluster config-
uration is this?
A. Active/spare
B. Active/passive

C. Passive/passive

D. Active/active

Copyright ©2002 SYBEX, Inc., Alameda, CA

166 Chapter 5  Fault Tolerance and Redundancy

12. RAID stands for which of the following?

A. Redoubtful Array of Inexpensive Diskettes

B. Redundant Array of Inexpensive Disks

C. Redundant Array of Independent Disks

D. A SWAT team action

13. Which of the following solutions combines data striping across disks
with mirroring?
A. RAID 1+5

B. Hybrid RAID 0+5

C. High Performance RAID

D. RAID 0+1

14. Your boss has asked you to implement hardware level RAID because
he understands that it is more reliable. He wants your opinion. What
will you tell him? (Select two.)
A. It is not necessarily more reliable but it does provide better
B. Software level RAID is less expensive but provides better perfor-
mance because the process is controlled by the network operating
C. Hardware RAID will cost more because he will have to purchase
a special controller and disks.
D. Software RAID will end up costing more because a special version
of the operating system has to be purchased.

15. How does RAID 5 write data to the disk?

A. In bits

B. In nibbles

C. In bytes
D. In binary

E. In blocks

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 167

16. Can a hot swappable disk also be a hot spare?

A. Yes.

B. No.

C. It depends on the level of RAID.

D. It depends on the speed of the SCSI devices.

17. You want a RAID solution for your server that will give you redun-
dancy in the event the disk itself or its controller fails. Which RAID
level will you choose?

B. Hybrid RAID 0+5


18. Your boss wants you to implement a level of RAID but he does not
want to incur any additional cost. You are running Windows 2000
on a server that has 4 physical hard disks. He understands that you
can set the RAID up in such a way that if one of those disks fails, the
data will not be lost. What are you going to implement? (Select all
that apply.)
A. Hardware level RAID

B. Software level RAID

C. Disk striping

D. Disk striping with parity

19. Your company is located in a remote area and the nearest vendor who
provides computer repair services is over 500 miles away. Which is the
best solution in a situation where a hard disk or a computer sustains
a fatal hardware failure?

Copyright ©2002 SYBEX, Inc., Alameda, CA

168 Chapter 5  Fault Tolerance and Redundancy

A. Implement a server with RAID 1 so that when a hard disk fails, it

won’t matter to the end users.
B. Keep hot spare disks on the shelf that match the type and config-
uration of the disks in your server. Then when a hard disk fails,
you can replace it yourself.
C. Entirely duplex your server.

D. Back up your data every night so you can still gain access to the
corporate data.

20. You wish to implement a RAID solution where data will be striped
across disks because you want the speed associated with data striping
on disk reads. However, you are concerned that if you lose multiple
disks, you will lose all the data on all disks. How can you protect your-
self against loss of multiple disks?
A. RAID 1+5

B. Hybrid RAID 0+5

C. High Performance RAID

D. RAID 0+1

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Review Questions 169

Answers to Review Questions

1. B. The minimum number of disks required to implement RAID level 5
is three. The parity information is spread out among the disks, and
disk space equivalent to one disk is used for parity.
2. A, C. Hardware level RAID does provide better performance
because the RAID controller has its own processor and cache. Soft-
ware level RAID has no special requirements while hardware level
RAID requires specialized hardware making it more costly.
3. C. RAID 0+1 (sometimes referred to as RAID 10) combines data
striping (RAID 0) and mirroring (RAID1). Data is striped across two
disks and mirrored onto another set of disks.
4. A. RAID 0, also known as disk striping, does not offer any level of
fault tolerance.
5. C. The smallest disk size is 10GB so the total volume size for the
array will be 40GB (10GB from each available disk).
6. D. There will be 30GB of space available for data storage and 10GB
will be used for parity information.
7. A. RAID 0 has the lowest disk overhead because there is no space
being used for parity information.
8. A, C. Software level RAID supports both SCSI and IDE disks in a
single array. It is also less costly because there is no need to purchase
specialized hardware such as a RAID controller.
9. C. In order to provide fault tolerance for the database application,
you should implement server clustering. If the server hosting the appli-
cation fails, another server can automatically assume its workload.
10. C. A cluster provides fault tolerance for a server and the applications,
services, and data it is hosting.
11. D. In an active/active cluster configuration, each cluster member
has its own workload so one server isn’t sitting idle while the other
one assumes all the work.
12. B, C. If you answered B, you are showing your age. When RAID
technology was first introduced, it was described as a Redundant
Array of Inexpensive Disks. Now, however, it is commonly called
Redundant Array of Independent Disks.

Copyright ©2002 SYBEX, Inc., Alameda, CA

170 Chapter 5  Fault Tolerance and Redundancy

13. D. RAID 0+1 is a hybrid approach where an entire stripe set without
parity is actually mirrored or duplexed.
14. A, C. Hardware RAID costs more because of the special controller
and disks that need to be purchased but it provides significantly better
performance than Software RAID. In addition it will support hot
swap of disks that fail.
15. E. RAID 5 data is striped at block level across all of the disks in
the chain.
16. A. Whether a disk is hot swappable has nothing to do with its status
as a hot spare.
17. C. Disk duplexing adds a second controller to the second disk, moving
the single point of failure away from the disk subsystem to the main-
board. In RAID 1, if either disk fails, the other disk takes over.
18. B, D. This is called disk striping with parity and can be implemented
via the network operating system. Therefore, it is software level RAID.
You do not need to purchase any additional hardware or software.
19. B. Hot spares can be replaced while the server is down for a minimal
time or the hot spares can be hot plug or hot swappable types.
20. D. RAID 0 is striping without parity, so it will give you the perfor-
mance you are looking for because it doesn’t have to calculate parity.
However, if you lose one disk, you will lose all data. If you use the
hybrid RAID 0+1, the 1 means that the disks will also be mirrored or
duplexed. Therefore even if you lose two or more disks, you will still
be able to get the data back from the mirrored disk.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Networking



 2.3 Install NOS

 Configure network and verify connectivity
 Verify network connectivity

 4.4 Perform physical housekeeping

 4.5 Perform hardware verification

 6.4 Identify and correct misconfigurations and/or upgrades

Please see Chapters 2, 4, 7, 8, and 10 for further

coverage of objectives on NOS installation,
physical housekeeping, hardware verification,
and identification/correction of misconfigura-
tions and/or upgrades.

Copyright ©2002 SYBEX, Inc., Alameda, CA

N etworks, quite simply, are groups of computers connected together to
share resources. This classic definition has stretched over the last few years
and is in need of refinement. Today the realm of networking has expanded
beyond the constraints of wires and simple file sharing to include PDAs and
wireless computer technology.
Networks are used to share resources. These resources can include files,
applications, printers, and Internet and mail connections. The benefits of
resource sharing extend beyond simple fiscal savings and into numerous
other areas. For example, many workers use the little yellow sticky notes
to leave messages for others. The chance of the sticky note becoming lost,
accidentally taken by someone else, or deliberately read by the wrong
person are high. A network facilitates reasonably secure communication
through e-mail, thus eliminating some sticky-note problems.
Network-based software that facilitates simultaneous use by multiple
users is also available. Several web-based products allow for such use
through a website; potential clients can use the website even while the
company is updating and/or redeveloping it. The time and cost savings
can be substantial. Network use has also spread into telephone conferences
and videoconferencing, bringing innovation to the way we conduct business.
This chapter will explain the current network technology, its benefits and
drawbacks, as well as the role of the server within the network environment.
As previously mentioned, the definition of a network has changed over
the last several years to include various new forms. What was originally
a small group of computers linked together with wires has now expanded
into a broad range of network types. Here I will classify seven:
LAN A LAN (local area network) is probably the most common form
of network in use today. LANs are often referred to as private networks.
A LAN is defined as a network located within the physical confines of a

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 173

geographical location (such as a building or group of buildings). LANs

have a limited distance of travel that is determined by the cable used.
MAN A MAN (metropolitan area network) is larger than a LAN but
not as large as a WAN. MANs consist of networks that span several build-
ings, city blocks, or even an entire city, providing redundant links and fast
access between these locations. An example of a MAN would be a net-
work between city libraries. Each library is a separate building located
within the geographical location of the city. Access to data between all of
these libraries can be gained through a MAN.
WAN A WAN (wide area network) is a network capable of spanning
large geographical and political boundaries. This includes states, coun-
tries, and continents. The Internet is the largest wide area network in
existence. It is an interconnection of computers worldwide.
SAN A SAN (storage area network) provides a link for multiple users
to a mass storage location. Large corporations that prefer to centralize
their data commonly use SANs. This assists in data safety and backups
because all the file servers can be housed in a secure building with limited
access. SANs normally use high-speed links such as fiber optics to
communicate with their storage facilities.
PAN A PAN (personal area network) is the newest network created.
A PAN is a small network that spans a very short distance (usually only a
few feet). These small networks allow for communication between PDAs
and desktop or laptop computers. PANs often use IR (infrared) or other
wireless technology to send and receive information.
VLAN A VLAN (virtual local area network) comprises computers and/
or resources that are physically on different network segments but com-
municate as though they were on the same segment. VLANs provide a
means of connectivity without the overhead of physically restructuring
several networks.
VPN A VPN (virtual private network) is a way of using public networks
to create private links. Most commonly it is a private connection made
through the public Internet infrastructure. The benefit of this is that the
users do not have to lease private lines or purchase expensive equipment.
A VPN is a secure tunnel created through an existing connection.
WLAN A WLAN (wireless local area network) eliminates the need
for cabling by using other communication media. These may be radio
frequency (RF), infrared (IR), or laser communication. The obvious

Copyright ©2002 SYBEX, Inc., Alameda, CA

174 Chapter 6  Networking

benefit of a wireless network is flexibility in structure. If you are not

limited by wires, the computing environment can take on almost any
form, shape, or location. Taking your laptop along on a nice day and
working outdoors, while maintaining network connectivity, is a reality.
Besides the types of networks available, there are also network topologies.
Topologies are the layout of the network. This can include both physical
layout (the way in which the network is actually installed and appears) and
logical design (the way in which information flows through the network).
Here we will classify five main topologies:
Bus In a bus topology all the computers are connected to one main cable
that runs the length of the network. This structure is often compared to a
string of Christmas lights. There is a beginning and an end, and between
these two points each computer connects to the wire. There are concerns
with the bus topology in that, much like a string of Christmas lights, if
there is a break in the cable or connection with a computer, then the entire
network can go down. Another problem with bus topology is locating
the fault when one occurs, because when the entire network fails, there
is no way to locate the fault besides starting at one end and working your
way to the other. This can be an extremely tedious—and often painful—
process, considering the network can span several floors and the wire is
often located within walls, drop ceilings, and crawl spaces under floors.
Figure 6.1 illustrates a bus topology.

FIGURE 6.1 Bus topology

Notice how there are special devices at the beginning and end of the bus
wire. These are actually 50-ohm terminators. Terminators are installed at
the beginning and the end of a bus network to eliminate signal bounce. If
a terminator is not present, is not working, or is not connected properly,

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 175

network signals will reach the end of the wire and bounce back (much like
an echo). With time, the entire bandwidth of the wire will be consumed
with signal bounce, thus bringing the network down. Bus connectors,
terminators, cabling, and cable standards will be discussed in more detail
later in this chapter.
Star A star topology is based around a central device such as a hub
or switch. Every computer is connected through its own cable directly
to this central device. This is an improvement over the bus topology, as
a cable failure will bring down only the one computer directly connected
to it. Star topologies are the most commonly used topology in small- to
medium-sized business environments today. Benefits of the star topology
are ease of installation and troubleshooting. Most central connectivity
devices have lights to indicate whether a cable segment is active or down.
Upgrading to add more devices to the network can be done without shut-
ting down the entire network. Figure 6.2 is an example of a star topology.

FIGURE 6.2 Star topology

A major concern with the star topology is the single point of failure
(SPOF). Having all network resources connected to one hub or switch
makes it a key component. Extra care should be taken in locating this
device in a safe and secure location. Another consideration is that, with
each device needing its own dedicated cable, there can be a multitude of
cables merging at the point of the central connectivity device. Cable
management becomes a focus. The cable used within a star topology
needs to be carefully routed throughout a building to avoid areas of
possible EMI (electromagnetic interference). As the number of cables
increases, this task can become increasingly difficult. Fortunately the
cost of cable used in a star topology has dropped, making it an afford-
able option. If longer cable runs need to be made to avoid possible EMI
interference, it will not be a major financial stress.

Copyright ©2002 SYBEX, Inc., Alameda, CA

176 Chapter 6  Networking

Ring The ring topology features all devices connected in a circular for-
mation. There are two different ring topologies to be aware of: logical and
physical. Logical ring topologies move information in a circular ring for-
mat but are physically a star topology. What this means is that there is a
central connectivity device and all other devices connect to this central
device thorough dedicated cables—it looks like a star topology. However,
information flows in a circular format throughout this network. Physical
ring topologies actually look like a ring. All networked devices are con-
nected in a circle. The advantage of a physical ring is that there is very
little cable in use, making installation easier. Since fiber optic networks
work on the principle of a ring, another possible advantage is speed. A
possible disadvantage of some ring topologies is that a single failure in
either the computer or in the cable can result in the entire network failing.

This is not the case in a true ring topology such as FDDI (fiber distributed data
interface). FDDI (described in more detail below in the “Fiber Optics” section)
“implements a dual ring so that it remains functional should one station die
or drop off the ring.”

Another disadvantage is that diagnosis of problems in a ring topology is

also difficult: Any service or updates on the network need to be done during
off-work hours because one system being shut down for service will halt the
entire network until the service is completed and the system is brought back
online. Figure 6.3 is an example of a physical ring topology.

FIGURE 6.3 Physical ring topology

Mesh A mesh topology is the most complicated and expensive topol-

ogy to integrate. However, it is the most resistant to failure from a cable
fault. In a mesh network every device is connected to every other device
through a dedicated cable. Should one cable fail, there will be other

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 177

paths that could be taken to reach the target resource. Although this
may seem like the ideal situation, mesh topologies can be a wiring night-
mare. The quantity and complexity of wires can become overwhelming.
Mesh topologies use by far the most cable and are the most complex to
install and troubleshoot. It is rare to see a mesh topology in use today.
Figure 6.4 is an example of a mesh topology.

FIGURE 6.4 Mesh topology

Hybrid A hybrid topology takes components of other topologies and

combines them in a mixed environment. The advantage of this is to use
the best features of each topology. For example, it is not uncommon to
see a mixture of a bus and a star topology. The bus structure is used as
a backbone link between servers within a building, and the star is used
to provide connectivity for all the other resources to the servers. Using
a high-speed cable option (such as fiber) as the backbone allows for fast
connectivity to the devices that will face the highest demands, such as
servers, while allowing for more affordable cabling between the end users.
Figure 6.5 is an example of a hybrid topology.

FIGURE 6.5 Hybrid Topology




Copyright ©2002 SYBEX, Inc., Alameda, CA

178 Chapter 6  Networking

Now that we have an understanding of the possible layouts of a net-

work, we can move on to explore the components of a network. Each
component within a network must meet specific industry standards. The
Server+ Exam will test your knowledge on the following standards.

IEEE 802 Standards

The IEEE (Institute of Electrical and Electronics Engineers) committee
created and released the 802 standards listed below to assist with industry
development of components as well as communications. When you consider
the complexity of components, as well as sheer quantity, there definitely is
a need for standards to maintain consistency as well as interoperability.
802.1 Bridging and Management Internetworking
802.2 Logical Link Control
802.4 Token Passing Bus Access Method
802.5 Token Passing Ring Method
802.6 DQDB Access Method, MANs
802.7 Broadband LAN
802.8 Fiber Optics
802.9 Isochronous LAN, Integrated voice/data networks
802.10 Network Security
802.11D Wireless Networks
802.12 Demand Priority Access
802.15 Working groups for WPANs
802.16 Wireless MAN
802.17 Resilient Packet Ring
Each of these standards focuses on a specific area of networking technology.
Full explanations of these standards as well as present and future developments
can be located at the IEEE website For the
Server+ Exam we will focus on the standards in common use.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 179

This standard defines a bus topology, using a 50-ohm coaxial baseband
cable with a transmission speed of 10Mbps. This was the original specific-
ation of Ethernet. It used CSMA/CD (Carrier Sense Multiple Access with
Collision Detection) to put data on the cable. CSMA/CD monitors the cable
for data traffic. When it senses that there is no traffic, it will attempt to send
data packets. If a collision occurs, then it will pause for a random period of
time and then attempt to retransmit. The problem with this type of network
is that the larger the number of clients and resources on the network, the
greater the number of collisions. This leads to slower network speeds.
Several new releases to this standard have emerged, providing new
cable options, connectors, and speeds of 100Mbps—and now 1000Mbps.
These newer standards of Ethernet will be discussed in detail later in this

Token Ring is another standard (based on the IBM PC Token Ring standard)
that has been frequently used. This standard specifies a physical star/logical
ring topology using twisted-pair wire. A special data carrier, called a token,
circulates through the ring from computer to computer picking up data
packets and delivering them to the destination. Each computer acts as a
repeater, boosting the signal of the token so it can travel to the next com-
puter. Only one computer can control the token at a time. When the data
packet and token reach their destination, the token unloads the packet
and takes a successful-reception acknowledgement to the sending computer.
If the sending computer has no more packets to transmit, the token then
becomes available to the network and the next computer waiting to trans-
mit. The advantage of this method of data transmission over the Ethernet
is that there are no collisions. Although there is only one token on the ring,
a Token Ring network can reach hundreds of systems and still perform

Fiber optics has taken on a strong role within the computer network envi-
ronment. Traditional hybrid networks used a coaxial cable backbone for
data transmission. This proved to be limiting in terms of transmission speed
as well as signal degeneration due to outside interference. Fiber cabling is

Copyright ©2002 SYBEX, Inc., Alameda, CA

180 Chapter 6  Networking

now actively replacing the coaxial cable as a backbone. The fiber allows
for faster transmission speeds, as well as immunity to EMI (electromagnetic
interference) and RFI (radio frequency interference). There are numerous
releases of fiber cabling today, because the best choices and applications are
still being identified. Fiber optics and fiber connectors will be discussed later
in this chapter.

Wireless technology has become the latest craze over the last few years. With
transmission speed and range increasing, wireless becomes an attractive
alternative to network environments that physically change on a regular
basis. Consider, for example, a network that needs to be set up in a historical
building, where drilling holes in walls and running cables is rarely an accept-
able practice. A wireless LAN is a much more acceptable option. Without
the installation of cables, a wireless network can be installed in a matter of
minutes. Wireless can also be integrated within an existing wired network to
provide a new breed of hybrid networks. This is often seen in environments
where laptop computers are used. Due to their nomadic nature, laptops can
remain connected to the network no matter where in the building they go.
Wireless technology is continuing to develop and improve. Handheld devices
also use this technology to communicate with host computers.
Wireless technology today uses an access point that sends and receives sig-
nals from the wireless devices. This access point in turn is wired to the net-
work (usually the switch or hub). Signaling methods on a wireless network
can include infrared, laser, narrow band radio, and spread spectrum radio.
Spread spectrum radio is often preferred to the other methods because the
price is reasonable but also because it does not require line of sight like
the laser technology does. Spread spectrum radio also can travel through
some walls, providing many options for the ever-changing network. Current
transmission speeds for wireless are in the 11Mbps range but steadily
increasing. Concerns with data security are also being addressed. If you are
broadcasting your network information over a radio frequency, it can be
captured by other devices. Wireless technology is still breaking new ground
in its development, both in speed and data security. With time, it will defi-
nitely become a major contender in the LAN arena. Right now though, it is
commonly used for small portable networks that regularly change location
and/or position.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 181

OSI Model
The Open System Interconnection Model is a theoretical seven-layer model
designed to illustrate the flow of information through a network. This
model was designed not only for aspiring network technicians to get a
better understanding of network communications, but also for technology
manufacturers to be able to dissect the elements and processes of informa-
tion flow and development, thus assisting in project development within
their own research departments. For the Server+ Exam you will need to
know the layers of the OSI Model as well as the functions that occur at
each layer.
Application Layer The application layer (layer 7) is at the top of the OSI
Model. At this layer, file and print services operate. This layer controls
data flow and error recovery.
Presentation Layer The presentation layer is responsible for the format
of data, network security, protocol conversion, data compression,
encryption, and translation.
Session Layer The session layer is responsible for establishing, main-
taining, and terminating communication sessions. These sessions are
often called virtual conversations. The session layer identifies passwords,
logons, network monitoring, and recovery from network failures.
Transport Layer The transport layer is responsible for error-free data
frames. It controls data flow and reliable end-to-end communication.
Network Layer This layer translates logical (TCP/IP) addresses into
physical (media access control, or MAC) addresses. The network layer
also determines the best path for information to travel on the network
if multiple paths exist.
Data Link Layer The data link layer is subdivided into two sublayers:
the MAC layer and the LLC (logical link control) layer. The data link
layer arranges data chunks into frames and organizes the frames into
a data stream, marking the beginning and end.
Physical Layer The physical layer, layer 1, is at the bottom of the OSI
Model. This layer describes how data is transmitted on the network cable
(media), including digital, optical, and mechanical interfaces.

Copyright ©2002 SYBEX, Inc., Alameda, CA

182 Chapter 6  Networking

The OSI Model is a complex interaction of layers. Information not only

travels up and down the layers but also horizontally, and directly with the
corresponding layer on the other machine. Each layer, as information flows
through the stack, adds or removes its own headers on the data packets. On
the sending computer, information begins at the top layer where the soft-
ware programs interact. The data is then sent down the layers until it reaches
the physical layer, where the data is sent out on the network wire. At the
receiving side the process occurs in reverse.

Of all the numerous standards and network structures, Ethernet has stood
out as the most popular network implementation. Ethernet has evolved
over the years to include several different cable types and topologies, as
previously mentioned. Through this evolution, improvements have been
made in reliability and speed.
Coaxial Based Ethernet The original Ethernet implementation was
based around the bus topology mentioned earlier. Although there were
several disadvantages to the original Ethernet (bus topology based), it was
affordable and easy to install. The cable of choice during this time was
coaxial cable. Coaxial cable consists of a center wire (usually copper)
surrounded by an inner layer of insulation, then a mesh or foil shielding,
and finally a thick outer PVC layer for protection. Figure 6.6 is an
example of coaxial cable.

FIGURE 6.6 Coaxial Cable

There are two common forms of coaxial cable in use with Ethernet:
Thicknet or 10Base5, and Thinnet or 10Base2. Thicknet cable is normally
used for a network backbone, as seen in the hybrid example (refer back to
Figure 6.5). This cable is very difficult to work with due to its thickness.
Many installers refer to it as the frozen garden hose. Trying to bend the cable

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 183

around curves can be as frustrating as rolling a frozen garden hose! Com-

puters are not connected directly to this cable; instead, sections of Thinnet
cable tap into the Thicknet cable and then run to the computers. The device
to attach these two cables together is called a vampire tap, which attaches to
the Thicknet wire though the use of a sharp spike that would pierce the PVC
shielding and come into contact with the inner copper core wire. Thicknet
cable can carry a signal for a maximum of 500 meters with a maximum
transmission speed of 10Mbps.

Cable Nomenclature

In the computer realm there has never been a shortage of acronyms, abbre-
viations, and epithets. Networking is no exception. Deciphering this jargon
can sometimes be an overwhelming task. In the example of 10Base2, key
information is presented in the name.

The 10 is the transfer speed, as measured in megabits per second (Mbps).

Transfer speeds have increased from the initial 10Mbps to 1,000Mbps.

Base stands for baseband signaling method. Baseband is a digital signal

over a single frequency. Broadband signaling is the other possible method.
Broadband signals are analog based and can send multiple signals simulta-
neously over the same wire using frequency division multiplexing (FDM)
or time division multiplexing (TDM). FDM and TDM are means of sharing
the same signal path by using the unique characteristics of analog signals
and either intermeshing where possible or sending one signal slightly
before another.

The 2 stands for an approximation of maximum allowable cable distance

represented in meters. Specifically, in a 10Base2 implementation the cable
cannot exceed 185 meters, but it is rounded off to 200 meters.

This makes sense until you reach 10BaseT. Now there is a letter instead of
a number representing the maximum cable distance. With this standard the
letter represents twisted-pair wiring. This includes both unshielded twisted-
pair and shielded twisted-pair (both of these will be discussed in detail later in
this chapter). Next comes the 100BaseF. The F, as you might have figured out
already, stands for fiber. This implementation would therefore provide
100Mbps transfer speed over baseband signaling with fiber cable.

Copyright ©2002 SYBEX, Inc., Alameda, CA

184 Chapter 6  Networking

Thinnet cable is much more flexible and is used for bus topologies and
cable runs from the backbone to the computer. Thinnet wire is classified as
RG58U cable. It can transfer data at a distance of 185 meters with a speed
of 10Mbps. Thinnet uses a BNC connector that attaches to each device on
the bus network. Figure 6.7 is an example of a BNC connector.

FIGURE 6.7 BNC connector

Depending on the location, a BNC tee connector may also need to be

used. This device allows for separate cable runs to connect at each side while
the center connects to the network card. At each end of a Thinnet bus there
must be a 50-ohm terminator installed. Figure 6.8 illustrates the proper com-
ponents in a Thinnet bus network.

FIGURE 6.8 Thinnet Network Components

Coaxial Cable

BNC T Connector
50-ohm Terminator

Network Card

Twisted-Pair Ethernet
Since that original implementation, Ethernet has expanded to include new
devices, cabling, and speeds. Current Ethernet is based on twisted-pair
cabling. Twisted-pair cable has opened the door to a whole new level of
Ethernet technology. Improved transmission speed and flexibility in instal-
lation, as well as new devices, have catapulted this media to the most popular
in use today. Figure 6.9 is an example of a section of twisted-pair media.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 185

FIGURE 6.9 Twisted-pair cable

Notice how the figure shows a solid-shaded wire and a striped wire
twisted together in a pair, and that there are four pairs of wires. The wires
are twisted together to help prevent signal interference. Signal interference,
either from another set of wires or from other devices (such as fluorescent
lighting) can affect performance on data wires. Twisted-pair cable is labeled
by category. Table 6.1 lists the common categories of twisted-pair cables and
their uses.

TABLE 6.1 Categories of Twisted-Pair Cable

Category Specifications

1 Voice only transmission, two twisted-pairs

2 Data grade at 4Mbps, four twisted-pairs

3 Data grade at 10Mbps, four twisted-pairs

4 Data grade at 16Mbps, four twisted-pairs

5 Data grade at 100Mbps, four twisted-pairs

5e Data grade at 1,000Mbps, four twisted-pairs

Twisted-pair is often referred to as UTP (unshielded twisted-pair). UTP is

the most commonly used twisted-pair wire, although STP (shielded twisted-
pair) is also available. STP is more expensive, but offers higher resistance
to electrical interference by offering a foil layer that surrounds the twisted-
pairs. The first UTP Ethernet transferred at speeds of 10Mbps and changed
the topology from a bus to a star. Later improvements increased the speed

Copyright ©2002 SYBEX, Inc., Alameda, CA

186 Chapter 6  Networking

to 100Mbps. In order to operate the network at 100Mbps, all the components

in the network must be supportive of the new speed. This includes cabling,
network cards, and connectivity devices. For a period of time, network
devices were released that supported dual speeds of 10 and 100.
Twisted-pair cable has a maximum transfer distance of 100 meters from
the computer to the connectivity device. When installing a UTP network, this
maximum cable distance includes all sections of wire—such as patch cables,
main cable runs, and wall drops. Patch cables connect computers to wall
plates, as well as patch panels to the switch or hub. The main cable run is the
actual cable that is located within the walls or floors. The drop is the section
of cable that comes down the wall or through a pipe and ends at a wall plate
or patch panel. The cable distance is calculated from the back of the com-
puter through all these points and ending at the switch or hub.
Due to the eight wires, UTP cables require a registered jack or RJ con-
nector. In telephone wire (which also uses UTP) an RJ-11 connector is used.
With network cables an RJ-45 connector is used. Figure 6.10 is a close-up
of an RJ-45 connector with UTP installed.

FIGURE 6.10 RJ-45 connector

Plastic Edge
Gold Connectors

Notice how the connector is pressed down on the outer layer of the UTP
cable. This process of installing the cable into the connector and securing it
is called crimping. Crimping requires a special tool called a crimper. The end
of the UTP cable is cut, and the PVC outer layer is trimmed back. Next, the
individual wires within the UTP are carefully rearranged into the proper
order (to match the standard being used). The wires are then slipped into the
RJ-45 connector, and the crimper forces small gold connectors into the ends
of the wire and also pinches a plastic edge onto the PVC jacket to hold the
end onto the wire. It is imperative that the plastic edge contacts the PVC
jacket and not the wire. Otherwise, the twisted-pairs could be damaged or,
over time, the end could slip off.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 187

STP connectors are identical to UTP connectors except for one added feature:
a metal shield. The exterior of the RJ-45 connector has a metal shield that
connects to the metal shielding of the STP wire. This ensures proper grounding
of the shielding throughout the entire length of the cable.

UTP cable can be crimped by either the A standard or the B standard.

The different standards determine the crimping order of the colored wires.
If you are crimping cable based on the A standard, make sure that your other
devices along the way are also based on the A standard. Table 6.2 lists
the A and B standards.

TABLE 6.2 UTP Cable Standards for Wire Crimping

A Standard B Standard

Green White Orange White

Green Orange

Orange White Green White

Blue Blue

Blue White Blue White

Orange Green

Brown White Brown White

Brown Brown

Network cable is often housed in plenum spaces, which are those areas
(usually above the ceiling or under the floor) used to circulate air. Running
cable in these areas poses a hazard in the event of fire because the cable can
give off toxic gases if it burns. Both twisted-pair and coaxial cable come in
plenum versions—these are coated with a fire-retardant material (usually

Copyright ©2002 SYBEX, Inc., Alameda, CA

188 Chapter 6  Networking

Most networks today are either running UTP cabling or changing from
coaxial to UTP cabling. The benefits of a UTP network, combined with the
ease of installation and troubleshooting, make it a smart choice.

Gigabit Ethernet
The latest Ethernet standard, which is becoming mainstreamed, is gigabit
Ethernet. These extremely fast implementations of Ethernet support data
transfer rate at 1,000Mbps. Currently hardware supportive of this standard
(including switches and network cards) is still very expensive. Implementa-
tion of gigabit Ethernet is often limited to backbones and server-to-server

Nearly all network cable sold today meets current fire and health codes.
However, some older cables may not. Before you upgrade or restructure
your network, be sure to check with your local building codes. Cable that
will be housed in plenum areas must be plenum-rated (must not release
toxic fumes when burned). Cable that is not plenum-rated is not acceptable
to use today.

Comparing the Ethernet Possibilities

Now that we have explored the various Ethernet cabling methods in detail,
a comparison of the essentials for each type may help clarify each standard.
Table 6.3 breaks down the essentials for each cabling type for you.

TABLE 6.3 Ethernet Cable Specifications

Cable Distance Nodes Speed

Thinnet 185 meters 50 10Mbps

Thicknet 500 meters 100 10Mbps

UTP/STP 100 meters 100 10–1,000Mbps

Fiber Optics
The latest technology is the use of fiber optic cabling. The benefits of fiber
cable include speed, transmission distance, and immunity to EMI and RFI.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 189

Initially fiber cabling was extremely difficult to install, requiring a trained

installer and very expensive equipment. This initial fiber cable was made out
of glass, and was therefore rather inflexible in installation. Today however,
improvements in plastics have made fiber cable more affordable and flexible.
Figure 6.11 is an example of fiber optic cable.

FIGURE 6.11 Fiber optic cable

Outer sheathing

Inner sheathing

Glass cable

Notice how the inner glass core is surrounded by several protective layers.
This is to ensure cable safety.
Fiber transmission methods fall under two categories: single mode and
multimode. Multimode uses light emitting diodes (LEDs) to create the signals.
The light produced by the LEDs contains various wavelengths. The diodes
shine light at the fiber, with some of the light wavelengths entering the fiber
and others not. The amount of light actually being used is not efficient and
therefore limits the distance that the signals can be sent. Multimode transmis-
sion fiber optics is the least expensive fiber option. Single mode transmission
relies on a laser as the light source. Laser light is one wavelength, and the
cable is matched with the laser to allow for maximum transmission. Due to
the purity of the laser light, single mode fiber implementations can achieve
transmission distances of 58 kilometers and are often used for long distance
Fiber cable is measured in microns. Fiber cable measurements are given
with two numbers—the first number is the diameter of the fiber strand and
the second number is the thickness of the cladding. These two numbers are
separated by a slash. The signaling method is also supplied. For example, a
62.5/12.5 multimode cable would contain a 62.5-micron fiber cable with
a 12.5-micron sheathing used only with a multimode signaling method.

Copyright ©2002 SYBEX, Inc., Alameda, CA

190 Chapter 6  Networking

Often fiber cable is bound together in multiple strands. This allows

for multiple signals and pathways to be used within one protective sheath.
Commonly there are between 2 and 24 strands, but there can be more.
Fiber connector types are plentiful. With a wide range of cable bundles as
well as interfaces, there have been numerous fiber connector designs. Two
of the most common are the FDDI (fiber distributed data interface) Media
Interface Connector and the SC multimode connector. The FDDI Media Inter-
face Connector is used on all fiber within the MMF-PMD and SMF-PMD
standards. Figure 6.12 is an example of a FDDI connector.

FIGURE 6.12 FDDI connector

The SC multimode connector is based on the LCF-PDM standards.

Figure 6.13 is an example of this style of connector.

FIGURE 6.13 SC connector

Network Devices
Besides topologies and cabling, there must be interface devices within a net-
work. These devices provide network connectivity. Each device has its own
specific use within LAN and WAN environments.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 191

Network Cards A network card provides a means of connectivity for

the computer to the network cable. The type of network card you select
must match the cable and network that you are using. A token ring net-
work card will not work on an Ethernet network, even though both use
the same cat5 cable. Even within the same network, such as Ethernet, care
must be taken when selecting a network card to ensure that it will match
the appropriate speed. If you have a 10Mbps Ethernet network, there is
little sense in purchasing a 100Mbps or 1,000Mbps network card, because
the network will not provide the high speed for you.

For more information on network adapter cards refer back to Chapter 5, “Fault
Tolerance and Redundancy.”

Repeaters Repeaters are used in an Ethernet network when the maxi-

mum cable length needs to be exceeded. A repeater will attempt to amplify
the signal and send it further down the cable. Unfortunately the repeater
will also amplify any noise or interference present on the cable. It is recom-
mended that the network be planned accordingly to eliminate the needing
for a repeater.
Hubs With the initial creation of the star Ethernet topology, the central
connectivity device was a hub. This was basically nothing more than a
multiport repeater. A passive hub simply allowed for connectivity between
all devices. Active hubs actually boosted signals and some even tried to
clean the signals before sending them out on the other ports.
Within a hub, signals are received on one port and then sent out all the
other ports. The hub neither reads the information nor discriminates as to
where the information goes. If information is traveling down a network
cable and encounters information traveling in the opposite direction at
the same time, collisions and resultant bandwidth compromise will slow
connections. In turn, slow network connectivity can lead to serious prob-
lems, including, for example, database crashes.
Patch Cables Patch cables are short cables (no less than one foot) that
are readily replaceable or interchangeable. They are used to connect a

Copyright ©2002 SYBEX, Inc., Alameda, CA

192 Chapter 6  Networking

patch panel to a hub or switch, a computer to a network wall plate, a

computer to a hub or switch, or other connectivity devices to each other.
Keystone Connector This is the female connector seen on wall plates
as well as the front side of the patch panel. Patch cables plug into the
keystone connector.
Patch Panels Patch panels are the central wiring point where twisted-
pair cables meet. Your network spans numerous computers across the
network, but each computer needs a dedicated cable that meets at the cen-
tral connectivity device. It is possible, but not advisable, to connect these
cable runs directly to your hub or switch. The potential for one or more
wires to be damaged puts these cables in risk. Should a cable be damaged,
it may have to be replaced, which would entail removing the entire cable
run (through the walls and floor or ceiling). This is a major project requir-
ing that the user connected to this cable run be offline for an extended
period of time. Patch panels connect to the main cable runs in the back,
and in the front use a short, replaceable patch cable to connect between
the patch cable and the switch or hub. Any cable damage normally occurs
as a result of cables being plugged and unplugged from the switch or hub.
Should a cable be damaged, it would be the small patch cable. This can
be easily replaced with a spare in a matter of seconds. Figure 6.14 is an
example of a patch panel.

FIGURE 6.14 Patch panel

RJ45 Ports

Racks Racks are becoming a common sight in networks. These devices

allow network hardware to be securely attached to a stationary object.
Standard racks are 19 in. wide and can have patch panels, switches, rout-
ers, gateways, servers, and cable management devices attached to them.
Attaching these expensive devices to something stationary makes a lot of
sense, unless you have money to spend on stolen devices. Racks that can
be completely enclosed are called cabinets, and those that are open to
airflow are open racks. Figure 6.15 is an open rack.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 193

FIGURE 6.15 Open rack

Bridge Bridges are interesting devices that allow you to virtually divide
your network into two separate LANs. Bridges can be used to connect
LAN segments that do not use the same media type. The benefit of imple-
menting a bridge is to decrease network traffic. When the bridge receives
a packet, it can determine which LAN segment the packet is destined for
and forward the packet to that segment and no others. It forwards packets
based on layer 2 addressing (MAC addresses). Bridges allow for specific
traffic to cross between the two virtual networks. For example, a printer
could be configured to work with both networks. Data packets addressed
to the printer will be acknowledged by the bridge and allowed to cross
over to the other network. In today’s networks, bridges are not often
used; switches and routers have become more common.
Switch A switch is often referred to as an intelligent hub. The benefit of
a switch over a hub is that a switch will read the information coming
inbound and, based on the address located in the data header, the switch
will send the information out on the receiving addressed port. This elim-
inates the network-wide propagation that occurs with a hub. Efficiency

Copyright ©2002 SYBEX, Inc., Alameda, CA

194 Chapter 6  Networking

with a switch is often the justification for the extra expense over a hub.
Switches are the connectivity device of choice today in both small and
large networks. Some high-end switches allow for monitoring data traffic
as well as creating VLANs; these are called layer 3 switches and are used
in high-traffic large networks.
Router Routers are intelligent devices that, if given multiple choices,
will select the best path for sending data between networks. WANs often
use routers to connect between locations. Routers can be complicated
devices to set up and maintain.
Routers are responsible for receiving packets and determining the best path
to reach the destination host using layer 3 addressing. Therefore, routers
can only be implemented if the LAN protocol (e.g., IP and IPX) supports
routing. It uses the logical addressing information (such as the TCP/IP
address) within the packet header to determine where to route the packet.
Routers maintain routing tables that contain information about destina-
tion networks. When a packet is received, the router uses the information
in the routing table to determine the best path to send the packet to reach
the destination network. The packet may be forwarded directly to the
destination network or to another router.
Gateway A gateway allows communication between different network
architectures and network environments. It will translate between proto-
cols and allow for dissimilar networks to be connected together and to

Piece of Cake

For years a local company provided high-end, low-profile surveillance

equipment to householders. The company managed both production and
business data on a small Linux-based LAN utilizing a star Ethernet topology
on a central hub. The LAN was administered by the business manager in her
spare time, but the company experienced few IT problems. Then business
really picked up, and what had been a small company suddenly found itself
doubling and redoubling its staff—and its network. Users started experi-
encing serious computing problems, including applications that timed out

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 195

and brought activity to a crashing halt. The business manager/server

administrator had more discretionary money than time, so she threw dollars
at the problem, purchasing more RAM, faster processors, bigger hard disks.
The problem didn’t go away. Baffled, she called in a consultant and hoped
for the best.

On the morning Joe. T. Consultant showed up, he found the network

down, the staff livid, and the business manager overwhelmed. When he
realized that 43 workstations were communicating through a passive hub,
the problem was as good as solved. A network topology that had been
appropriate for 10 users couldn’t stand up to 43. Data collisions at the hub
were wreaking havoc on the entire network. The solution? A switch was
ordered, a generous installation fee agreed upon, and Joe was home in
time for dessert.

Network Installation
Now that we have a clear understanding of each network component, we
can start putting it all together. There is a logical process of steps from the
planning stage to the finished, up-and-running network.
To begin with, you have to determine which network is going to best
suit your needs. How large is your network? What are the software and
hardware uses of the network? Where will the computers be located? What
is the maximum distance that your network will need to stretch to? Will the
network grow with time? What are the future plans for the business in terms
of software and hardware requirements?
Most small- to medium-sized networks today will install an Ethernet net-
work based on a star topology with a switch running at 100Mbps. This
network will use category five UTP cable and might or might not have a
patch panel attached to a rack. Begin with deciding on the server’s location.
This might be a dedicated room, closet space, someone’s office in the
back corner, or other possible locations. As you will learn in Chapter 13,
“Managing and Securing the Server Environment,” there are numerous
environmental variables to take into consideration when selecting the loca-
tion of your server and networking equipment. Normally the connectivity
devices and server are located within the same setting. The ideal environment
would be within a rack, both for security and to provide a stable mounting

Copyright ©2002 SYBEX, Inc., Alameda, CA

196 Chapter 6  Networking

The first step is the wiring. If you are replacing an existing network, you
can often follow the same wire paths that were in use before. However, if
you are installing a new network, then care must be taken when running the
main cable lengths to ensure that they are routed away from sources of EMI
or RFI. Care when crossing electrical wires must also be taken. It is advisable
to cross electrical wires at 90-degree angles rather than running the UTP
parallel to the electrical wire. When running your network wires, be sure to
label each wire with its source location. Remember that at the switch end
there will be a wire for each computer. If you do not label each wire with its
source, you will have no idea which wire belongs to which location. Special
numbered stickers and tape are available, but they can easily fall off. A
permanent felt marker ensures that your labeling system will not become lost
over time. In small networks that don’t change, you can label each wire with
the user’s name. However, in a larger network or an office that changes staff
regularly, this system is not advisable—for obvious reasons.
Once all wire has been run, the ends can be crimped or attached to the
keystone connectors. It is highly advisable to attach keystone connectors and
wall plates to the UTP at the user’s computer end. This not only creates
a much cleaner installation, but also protects the connection: The cable
between the wall and the back of the user’s computer can, and often does,
take abuse. Becoming tangled in feet under the desk, tension from the com-
puter being moved, and danger from cleaning staff and vacuums are just a
few of the possible dangers that this cable can encounter.
By installing the wall plate and keystone connector, you then use a short
patch cable from the wall plate to the back of the computer. If the cable
becomes damaged, it can be readily replaced with a spare cable, with no
stress or damage to the main cable run. At this time, the UTP can also be
attached to the back of the patch panel. These cables, just like at the keystone
connector side, are attached with a punch down tool. This tool is really a thin
blade that forces the strands of cable, by their color codes, into small gold
connectors that provide the connectivity. Once completed, the labels for the
wires can be transferred to the front of the patch panel and the patch panel
can be installed into the rack or on a wall to ensure that there is no further
stress on the main cable runs.
Now that all of the cables have been run, the next step is to test the cables.
Cable testers come in a variety of forms. The simplest ones test for basic con-
nectivity. A transmitter connects at one side of the UTP through the RJ-45
connector and a receiver connects at the other end. The transmitters send

Copyright ©2002 SYBEX, Inc., Alameda, CA

Networks 197

electrical pulses down each strand of cable within the UTP and the receiver
lights for each wire. If there is a fault with crimping, you can determine
which wire is failing. The most expensive testers will provide the same
function but also test for signal degradation, speed of transfer, and numer-
ous other elements; these testers will also provide a printout of the test
results. Since these testers are expensive, they are often rented.

When you are testing network cables, ensure that there are no devices
attached to the other end of the cable. The electrical impulses sent by the
testers can transmit enough charge to damage your expensive network

Once the cables have been installed and tested, the connectivity device can
be installed. Depending on your budget and network size, you will have to
choose either a hub or a switch. As previously mentioned, hubs will propa-
gate information through every wire on the network and can lead to slow
network performance. If your network has more than 12 users and tends
to send and receive a large amount of network traffic, then you will want to
purchase a switch. Whether you buy a hub or a switch, you must select
carefully. Many hubs and switches allow for stacking. Stacking is the ability
to link devices together. So, should your network grow beyond the number
of available ports on your switch or hub, you can simply buy another hub or
switch and link it to the first one. However, when purchasing the initial hub
or switch, you should have a clear idea of the number of current and future
computers that will be connecting to the network. Then purchase a hub or
switch that will meet this need.
Switches and hubs commonly come with 4, 8, 16, or 24 ports. If you need
13 ports, then you will have to purchase a 16-port switch. Once you have
decided on a switch or hub that will meet your needs, it can be installed into
the network. If you are using a rack system, the switch or hub will be
installed below the patch panel. This will ensure that the wires from the
patch panel to the switch will not interfere with other devices installed in
the rack. When installing any components in the rack, care should be taken
with electrostatic discharge. ESD can cause damage to components even if
you don’t open the case. Use an ESD wrist strap or other suitable ground
methods to ensure that you are not going to cause a potential fault in your
new components.

Copyright ©2002 SYBEX, Inc., Alameda, CA

198 Chapter 6  Networking

Finally the network cards can be installed and configured. If your com-
puter does not have a network card installed, the first step is to perform a
safe shutdown of the server and take the cover off the computer and examine
the available expansion busses. Hopefully you have a free PCI slot and can
then purchase a PCI network card. Since you are installing a 100Mbps net-
work, the card would have to support the 100Mbps transfer speed. While
wearing an ESD wrist strap, you can then install the PCI network card. If you
are using a PCI card and Windows, the card should be Plug and Play so that
the operating system will identify the network card when you turn the com-
puter back on. Insert the manufacturer-supplied disk when prompted and
the installation of the network card will be completed.
Connect patch cables between the computers and the wall plates, and
between the patch panel and switch. (Remember to test your patch cables
before installing them.) Connect patch cables to the servers network cards
and the switch.
The network hardware is now installed. Next comes the operating system
and software configuration. This will be covered in detail in Chapter 7,
“Network Operating Systems,” and Chapter 8, “TCP/IP.”

T his chapter began with an exploration of major network types. The
most common types of networks seen today include the LAN, WAN, SAN,
VLAN, VPN, and WLAN. Network topologies are the physical layouts that
a network can have. This includes the bus, star, ring, mesh, and hybrid.
Networks are governed by standards to ensure compatibility between
vendors. IEEE created the 802 standards to subdivide the various network
areas into 17 different areas. Of these the most commonly implemented ones
are 802.3 Ethernet, 802.5 Token Ring, 802.8 Fiber Optics, 802.11 Wireless.
The OSI Model is a layered approach to understanding the flow of infor-
mation through a network. The layers from top to bottom are Application,
Presentation, Session, Transport, Network, Data Link, and Physical.
Of the various networks available, Ethernet is the most common imple-
mentation in use today. Initially a bus topology using coaxial cable and
transmitting at 10Mbps, Ethernet has grown to a star topology and uses

Copyright ©2002 SYBEX, Inc., Alameda, CA

Exam Essentials 199

UTP cabling at speeds of 100Mbps. UTP cable is available in six different

categories; today, category 5 cable is the most widely used and can transmit
data at 100Mbps. It uses four pairs of copper wires. The next generation of
Ethernet will use category 5e cable and transmit at 1,000Mbps. This is called
gigabit Ethernet.
Fiber optics fall under two categories: multimode and single mode.
Multimode fiber optic systems use LEDs to transmit light through the cable.
Single mode fiber optic systems use a laser as the light source, which is more
efficient and capable of transmission of greater distances than multimode.
Network devices include all the common components used within a net-
work environment. This can include network cards, repeaters, hubs, switches,
routers, gateways, bridges, patch panels, patch cables, racks, and keystone

Exam Essentials
Know the different types of networks. Be able to identify the differ-
ences between LAN, WAN, and MAN networks.
Know the different network topologies. Identify and understand
the benefits and drawbacks of the bus, star, ring, mesh, and hybrid
Be able to identify the IEEE 802 standards. Pay special attention to
802.3, 802.5, 802.8, and 802.11. Know the details of each of these four
levels. Be familiar with identifying all the 802 standards by function.
Know the layers of the OSI Model. Be able to list, in the proper order,
the layers and functions of the OSI Model.
Know the Ethernet standards. Make sure to understand the differences
in Ethernet evolution from coaxial bus topologies to UTP star topologies.
Understand how a bus network works. Be familiar with Thinnet,
Thicknet, backbone cables, terminators, tee connectors, and the bus
Understand how a star network works. Be familiar with UTP, STP,
central connectivity devices, crimping, and patch cables.

Copyright ©2002 SYBEX, Inc., Alameda, CA

200 Chapter 6  Networking

Know the categories of UTP cable. Be able to identify the categories by

speed, pairs of wires, and use.
Know the different fiber optic signaling methods. Be able to decipher
the differences between multimode and single mode fiber optic signaling.
Know the functions of common network devices. Identify and under-
stand the uses of hubs, switches, repeaters, brides, routers, gateways,
patch panels, racks, keystone connectors, and patch cables.

Key Terms
B efore you take the exam, be certain you are familiar with the follow-
ing terms:

10Base2 OSI Model

10Base5 PAN (personal area network)
10BaseT ring topology
802.11 RJ-45
802.3 SAN (storage area network)
802.5 single mode
802.8 star topology
bus topology switch
hybrid topology UTP (unshielded twisted-pair)
LAN (local area network) VLAN (virtual local area network)
MAN (metropolitan area VPN (virtual private network)
mesh topology WAN (wide area network)
multimode WAN (wireless local area network)
network card

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 201

Review Questions
1. What are networks?

A. Groups of computers to share resources

B. Connections between computer parts

C. Links between printers

D. Groups of computer components for a user

2. Which of the following is not a commonly shared resource?

A. Monitors

B. Files

C. Printers
D. Internet connections

3. Which of the following is physically the largest type of network?





4. What will happen if a device fails on a bus topology?

A. Nothing.

B. That device will not work.

C. The entire network will fail.

D. Part of the network will fail.

5. What will happen if the central connectivity device fails on a star

A. The entire network will fail.

B. Only the central device will fail.

C. Nothing.

Copyright ©2002 SYBEX, Inc., Alameda, CA

202 Chapter 6  Networking

6. What topology uses the most amount of cable?

A. Star

B. Ring

C. Bus
D. Mesh

7. Which 802 standard is Ethernet?

A. 802.5
B. 802.11

C. 802.3

D. 802.7

8. Which 802 standard is Token Ring?

A. 802.5

B. 802.3

C. 802.11

D. 802.7

9. Which 802 standard is wireless technology?

A. 802.3

B. 802.5
C. 802.7

D. 802.11

10. Which layer of the OSI Model is responsible for logical (TCP/IP)
A. Network
B. Transport

C. Physical

D. Data link

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 203

11. Which layer of the OSI Model is responsible for describing how data
is sent on the network media?
A. Transport

B. Physical
C. Network

D. Session

12. Thinnet cable is also referred to as which of the following?

A. 10Base5

B. 10BaseF

C. 10BaseT

D. 10Base2

13. Thicknet cable can transmit data to what distance?

A. 100 meters

B. 185 meters

C. 200 meters

D. 500 meters

14. In category 5 UTP, how many pairs of wires are there?

A. Two

B. Four
C. Six

D. Eight

15. Category 5 UTP can transmit at what maximum speed?

A. 10Mbps
B. 10,000Mbps

C. 1,000Mbps

D. 16Mbps

Copyright ©2002 SYBEX, Inc., Alameda, CA

204 Chapter 6  Networking

16. UTP can transmit at what maximum distance?

A. 100 meters

B. 185 meters

C. 200 meters
D. 500 meters

17. What light source does multimode fiber optics use?

A. Laser

C. Natural light

D. Impulse

18. What is the light source for single mode fiber optics?
A. Laser


C. Natural light

D. Impulse

19. What is the benefit of a switch over a hub?

A. Switches have fewer ports.

B. Switches transmit data to the receiving port only.

C. Switches propagate incoming data to all outbound ports.

D. Switches are more affordable.

20. Which of the following network connectivity devices uses layer 3

A. Hubs
B. Bridges

C. Switches

D. Routers

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Review Questions 205

Answers to Review Questions

1. A. Networks are groups of computers connected together to share
2. A. Monitors are not a shared network resource.
3. B. WANs are the physically largest network.

4. C. On a bus topology a single device failure will halt the entire

5. A. Trick question. If the central device fails, then the entire network
will not be able to communicate.
6. D. Mesh topologies use the most cable.
7. C. The Ethernet standard is 802.3.

8. A. The Token Ring standard is 802.5.

9. D. The wireless standard is 802.11.

10. A. The network layer is responsible for logical network addressing.

11. B. The physical layer is responsible for network media.

12. D. Thinnet is also called 10Base2.

13. D. Thicknet can transmit up to 500 meters.

14. B. There are four pairs of wires in category 5 UTP.

15. C. The maximum transmission speed of category 5 UTP is 1,000Mbps.

16. A. UTP can transmit at a maximum distance of 100 meters.

17. B. Multimode fiber uses LED as the light source.

18. A. Single mode fiber uses a laser as the light source.

19. B. Switches will transmit data only to the receiving port.

20. D. Routers use layer 3 addressing to route packets between network


Copyright ©2002 SYBEX, Inc., Alameda, CA

Software PART

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Network Operating

 1.1 Conduct pre-installation planning activities

 Plan the installation
 Verify the installation plan
 Verify hardware compatibility with operating system
 Verify power sources, space, UPS and network availability
 Verify that all correct components and cables have been

 2.3 Install NOS

 Configure network and verify network connectivity
 Verify network connectivity

 2.4 Configure external peripherals (UPS, external drive

subsystems, etc.)

 2.5 Install NOS updates to design specifications

 2.6 Update manufacturer specific drivers

Please see Chapter 9, “Upgrading and Mainte-

nance,” and Chapter 10, “Hardware Updates,”
for further coverage of the first exam objective
above: “Conduct pre-installation planning

Copyright ©2002 SYBEX, Inc., Alameda, CA

I n this chapter you will learn about a number of common operating
systems available for use on network servers. CompTIA seems to have a
relatively uncomfortable relationship with this part of the Server+ Exam,
because their policy is to strive for “vendor neutrality” but this is pretty
much impossible when dealing with server software. As such, the exam
instead incorporates questions that test a candidate’s knowledge of the basic
functionality and commands of a number of different operating systems.
In this chapter the three main network operating systems (NOSs) that are
specifically named in the exam objectives will be dealt with individually,
allowing you to get a very basic sampling of the tools and options available
in each. Others will be mentioned in name and briefly described to inform
you of the options, but the primary emphasis will be on these:
A number of these operating systems have multiple versions/flavors/dis-
tributions, and those will be discussed later, but because we are not going
very deep into any of these operating systems (large books have been written
on each of them), only factors that are common to all versions of the OS, and
of importance to the Server+ exam, are likely to be considered.

NOS Options
W hen getting ready to purchase any software for your server, a
number of considerations come into play. When purchasing the NOS,
these choices become even more complex.
In choosing your network OS, you are also determining which applica-
tions you will be able to purchase, how your network will be managed, and

Copyright ©2002 SYBEX, Inc., Alameda, CA

NOS Options 211

what types of advanced functionality will be available to you. Because of

this, it is important that you understand the basic advantages and disadvan-
tages of each NOS available.
First off, there is one simple but true rule that you should always keep
in mind: Whatever your needs, it is generally best if possible to stay with a
single hardware and software vendor for your servers. There are cases, of
course, where particular applications only run on Windows 2000, and so to
use that app you will need to add a Windows 2000 server to your NetWare
network. Or you may be migrating from OS/2 to a Linux environment, but
one application you need does not have a Linux equivalent, so at least one
OS/2 server will survive the changeover.
When designing a new network server environment, though, it may be
best to look for the NOS that best fits your overall network needs, and then
try to plan out a network structure that allows for most—if not all—of
your network servers to run on a consistent platform. This makes training,
upgrades, security, bug fixes, and a host of other administrative tasks
easier because there is a smaller range of documentation and changes
to keep up with.
This sort of practical info may go against the “feel” of CompTIA
vendor-neutrality, so there is the possibility they will disagree and recom-
mend that you use the “best OS for each job.” That is the problem with
questions that have no “right” answers. Because the Server+ Exam is
based on the consensus of a number of SMEs (subject matter experts), and
because even experts are rarely unanimous on how to do any computer
task, it is difficult to know for certain what the correct choice is.

In this book, we will present you with the data needed to answer the factual
elements of the exam, and we will also give you the mental framework from
within which you can make the logical decisions that are such a big part of the
test. As you answer the questions, though, remember that the question is not
asking you how you think something should be done—it is asking how you
think the majority of experts would recommend it be done.

The following is a basic list of things to consider when choosing a network

operating system:
 Application compatibility
 Hardware requirements

Copyright ©2002 SYBEX, Inc., Alameda, CA

212 Chapter 7  Network Operating Systems

Application Compatibility
Application compatibility concerns are generally pretty straightforward.
Usually a particular software package is either written to run on a particular
OS, or it isn’t. Even so, there are times when the software is only compatible
with certain versions of a NOS, or where the application is more full-
featured on one platform than on another.
The simple fact is that the easy way out here is to pick Windows 2000
Server as your NOS, because just about anything you could need is written
for the Windows OS family. That doesn’t make Windows the automatic
choice for compatibility, though. At times, security, stability, or other key
elements of an operating system influence your choice. For example, e-mail
servers often use a Unix flavor for their operating system. With the mul-
titude of viruses on the Internet, and e-mail being the most common
transport method for a virus, Unix operating systems provide the best
tolerance to virus infections.
Other applications are designed to operate within a specific operating
system environment. This is in part due to resource management, but more
commonly due to shared files such as drivers. The best way to eliminate
possible problems is to clearly research the programs and software that you
are installing to ensure that they will support installation on the operating
system you are using as well as the version. Each NOS manufacturer has
released several versions and update patches. Each change will result in an
impact on the software that you wish to run on the server. This also raises a
concern when you decide to perform an operating system upgrade or patch.
Consideration of the applications installed must be taken.

Hardware Requirements
Not all operating systems are the same—most obviously in their installation
and user interface, but also in their hardware requirements. Hardware
requirements include both physical hardware installed within the server as
well as available resources for operation (such as RAM and virtual memory).
Selecting an operating system that will best meet the hardware requirements
(or vice versa) is an important planning step.

Novell NetWare
The latest Novell NetWare release is NetWare 6. The following hardware
requirements are the minimum installation and operation requirements for
NetWare 6:
 Intel Pentium II or higher processor

Copyright ©2002 SYBEX, Inc., Alameda, CA

NOS Options 213

 VGA or higher resolution display

 DOS partition of at least 200MB with 2GB System volume
(4GB recommended)
 256MB RAM (512 recommended)
 CD-ROM drive
Novell operating systems have traditionally run on minimal RAM
requirements, due in part to the way in which devices are installed and
configured through NLMs. Netware Loadable Modules allow you to
install only the components that are needed. This frees up space and
resource use tremendously.

More information on the newly released Novell NetWare 6 can be

found at

Windows NT
Windows NT preceded Windows 2000. Hardware requirements between
the two operating systems remained similar, with Windows 2000 requiring
a bit more in the resource area. The new .NET server (Windows Whistler),
due to be released later in this year, will continue the same trend. The fol-
lowing requirements should continue to serve:
 Intel 80486 processor or higher
 VGA display
 125MB free hard drive space
 At least one network card

Copyright ©2002 SYBEX, Inc., Alameda, CA

214 Chapter 7  Network Operating Systems

Windows 2000 Server

As previously mentioned, Windows 2000 hardware requirements are similar
to those of Windows NT. The obvious change occurs in the amount of hard
drive and RAM required:
 Intel Pentium 133MHz or higher
 128MB RAM (minimum) recommended (4GB maximum)
 1GB available hard disk space
 VGA or higher-resolution monitor
 Network card

More information on Windows 2000 can be found at


With the multitude of flavors of Unix available, it is impossible to give a clear
list of resources that must be met to have the Unix-based operating system
function on a server. With the source code for Unix open to public alteration
and tweaking, the variations of this operating system increase and grow on
a regular basis. The following is a general hardware requirement list for
Mandrake Linux (one of the many Unix distributions).
 Intel Pentium processor
 500MB hard drive space

More information on Linux can be found at

Copyright ©2002 SYBEX, Inc., Alameda, CA

NOS Options 215

OS/2 Warp
OS/2 Warp is a server operating system designed by IBM for their server
line. It is a scalable OS with support for numerous business solution prod-
ucts. In the words of IBM, “IBM solutions target today’s heterogeneous,
open computing environment.” This server solution is often installed and
configured upon arrival from IBM and requires little further configuration.

For more information on OS/2 Warp go to


Features of an operating system are often a major element in making your
decision to purchase one OS over another. In the past, the features were
dramatically different between the operating systems. Windows NT came
with a graphic user interface, while older versions of Unix and NetWare 3
did not—they were command-line-based, much like old DOS. This feature in
itself led to increased sales for Microsoft because the GUI eased the daily
chores of installing users, printers, mapped paths, and security.
Other features to consider include ease of installation and setup. Will
there be a steep learning curve involved in implementing this server opera-
ting system? What are the desired uses for the server on the network? We
previously mentioned the e-mail server and e-mail viruses. If the server is
to be an e-mail server, then the use of a Microsoft server operating system,
even with virus protection, is still a risk.
Another area of consideration is interoperability with the rest of the
operating systems within the network. What are the clients using, both
as hardware/operating systems and also software? Will it be compatible
with the new server operating system? Will there need to be adjustments
made to the clients’ computers? Will this server operate in an environment
with other servers using different NOSs? What network protocols will be
used as communication between devices on the network as a result of the
server operating system selected?
Windows historically has relied on NetBEUI as its protocol of choice
(up until Windows 2000), Novell has favored IPX/SPX (until NetWare 5),
and Unix prefers TCP/IP. Today all three manufacturers have agreed upon
using TCP/IP, but you will probably work in a heterogeneous environment
because it is a rare company that has the fiscal resources to purchase the

Copyright ©2002 SYBEX, Inc., Alameda, CA

216 Chapter 7  Network Operating Systems

latest hardware and software as soon as it comes out. This leaves you often-
times in an environment with dated Novell or Microsoft servers, or both
operating together.
Key features of each operating system will be discussed within the next
few sections. The thing to remember is that an assessment of your network
needs will reflect heavily on the operating system and version that you
decide to use.

Cost is always a concern. As much as we would like to have the best of every-
thing in the world, the reality is that few of us can financially do it. Server
operating systems will range in price dramatically as the issues of licensing is
brought up.
Unix/Linux is by far the least expensive operating system. In many circles
it is free for distribution or a nominal fee is charged. Due to its open archi-
tecture, it can then be reengineered to best meet your business needs. The
concern with Unix, at times, has been support for third-party drivers. You
might not be able to locate a driver for your video card or network card, for
example. Driver manufacturers, as of late, have seen the demands and trends
toward using Unix as a common operating system and have been busy
playing catchup, designing Unix-specific drivers.
Windows servers come in several forms. Each was designed to meet a
specific business need. Windows NT Server could be purchased as well as
Windows NT BackOffice Server 4.5; the latter provided an entire suite of
software programs that were designed to control everything from Internet
proxy to database sharing. Windows 2000 also offers a few different
options, including Windows 2000 Server and Windows 2000 Advanced
Server. Windows 2000 has been rather expensive; current price for Win-
dows 2000 Server and a five-client license is $1,344.99. The expense
increases dramatically with licensing additional users. Not only do you
have to pay for the operating system but a license fee must also be paid
for each computer that will be connecting to the server.
Before purchasing client access licenses, you must determine which
licensing method your server will use. The two choices are Per Seat and
Per Server. You can only use one licensing method in the server. If you
choose Per Server, you can switch to Per Seat at some point in the future,
but once you are licensed Per Seat, the decision is permanent. Microsoft

Copyright ©2002 SYBEX, Inc., Alameda, CA

NOS Options 217

recommends that if you are unsure of which mode to choose, you should
choose Per Server, since that allows you the flexibility to change modes
later on.
Licensing Per Seat means that it is the client end of the connection that
holds the license, and that license can be used to connect to any server on the
network. Licensing Per Server means that it is the server end of the connec-
tion that holds the license, and there must be an available license in order
for each client to connect. There are two factors that tell you which is the
cheapest option for your server:
 Number of servers on the network
 Number of concurrent connections that clients will make to the
If you have only one server on your network, it will most likely be best to
choose Per Server, because you will only have to purchase enough licenses
to equal the number of concurrent client connections. In this scenario, you
could potentially have many fewer licenses than client PCs. This is particu-
larly true if the clients connect to the server, use the server resources, and
quickly disconnect. If client PCs will maintain a connection for a long time,
then the number of licenses will probably equal the number of clients, which
is the same cost as licensing Per Seat in the case of having only one server. An
example where this strategy would work well would be in a remote access
server. If clients are connecting into the server remotely to check their mail
(for example) and then disconnecting, there is no need for a license for every
computer: Per Server is the best option.
If you have more than one server on your network, it will most likely be
best to choose Per Seat. The only way it wouldn’t be the best way is in the
peculiar case of having very few concurrent client connections. When licens-
ing Per Server, each server contains a pool of licenses, so that if one server
has 25 licenses, and another server has 10 licenses, you can only legally
connect 25 clients to the first server and 10 to the second. In this same
situation, if you licensed Per Seat, you could have 35 clients connect to
either server or both servers simultaneously.
In the real world, there are very few situations that will warrant Per Server
licensing. Only if you have a single server or very brief server connections
will Per Server be cost effective. In contrast, Per Seat licensing allows for
client PCs to connect to as many servers as are available on the network, with
no thought for other clients’ concurrent connections.

Copyright ©2002 SYBEX, Inc., Alameda, CA

218 Chapter 7  Network Operating Systems

Major Network Operating Systems

As we previously discussed, three major operating systems are in
extensive use today. Although a multitude of others are available, those from
Microsoft, Novell, and Unix distributors have really dominated the field.
Throughout this next section we will explore the major releases, features,
and tools of each one.

Novell NetWare
Originally known as Share Net, Novell’s NetWare NOS has been evolving
for nearly 20 years, making it the dean of the three NOS platforms we will
be discussing.
Share Net debuted in 1983, and became NetWare shortly after. Early ver-
sions of NetWare were extremely successful in competing with Microsoft’s
LAN Manager and Banyan Vines. Both of these products have followed the
evolve or die motto, with LAN Manager evolving into Windows NT/2000,
and Banyan Vines dying.
NetWare has a bit of both of these possibilities in its past. In the mid
1990s it seemed as though NetWare was everywhere (most estimates showed
that over 80 percent of all LANs ran on NetWare in 1995). When Novell
brought out NetWare 5.x with a distributed network directory based on
the x.500 standard, they appeared certain to crush all other competitors.
Their NOS was better technologically than any of their competitors’, they
had terrific market share already, and their customers were dedicated to the
company and the product.
Oddly, that was about the time the wheels came off.
Four things seem to have occurred, more or less at the same time, that
caused Novell serious problems:
1. The Web started reaching corporate networks around 1995.

2. Microsoft brought out Windows NT 4 in 1996.

3. Novell purchased WordPerfect, with the intention of matching

Microsoft with an OS/Productivity suite pairing.
4. The hardware explosion began. PCs went from 486/66MHz in 1995
to Pentium 4/2GHz as of Fall 2001.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Major Network Operating Systems 219

In short, here is the reading of one analyst about how this affected the
company: Even as the most important technological revolution of the decade
was going on around them, Novell spent its resources developing a word pro-
cessor, and neglected to market their superior NOS. Microsoft, meanwhile,
took advantage of the fact that faster hardware allowed their GUI servers
to compete (somewhat) with Novell, and hit the advertising/marketing trail
hard for NT. Microsoft quickly standardized on TCP/IP, and rode the
wave of the Internet. NetWare did not switch to all TCP/IP for another
four-plus years.
Because of this, much of Novell’s market share has evaporated, and the
company has a number of bridges to repair. Even so, there are significant
reasons for optimism. They still have a great NOS, they have dumped
WordPerfect, and they have a new management team.
Here are some of the specs you will want to know about NetWare for
the exam. These are generally written with an eye toward giving you “just
enough NetWare.” The same practice will be followed for the other OSs.
If you want detailed knowledge, check out the links and books mentioned.

Major Versions of Novell NetWare

You will not encounter a multitude of NetWare versions—either on the
Server+ Exam or in the real world. Novell (unlike Microsoft) has not
released a dizzying array of products over the years.

NetWare 3.x
NetWare 3.x included NetWare 3.11 and 3.12, based around the product
known as NetWare 386 having been introduced concurrently with Intel’s
386 chip. NetWare 3 supported multiple cross-platform clients (Microsoft,
Apple) and had minimal hardware requirements (4MB RAM, 74MB hard
disk space); this allowed for NetWare to be installed in low-cost environ-
ments. NetWare 3.x used a database called the Bindery to maintain groups’
and users’ accounts. The three major utility programs used (through a
command line interface) to control a NetWare 3.x server were Syscon,
Syscon was used for user administration of the Bindery.
PCONSOLE was used for printer setup.
FILER was used for file operations.

Copyright ©2002 SYBEX, Inc., Alameda, CA

220 Chapter 7  Network Operating Systems

The simplicity and stability of NetWare 3.x made it a favorite among

network administrators. The downside was that you had to switch between
each of the three utility programs to perform all the required tasks. This
meant exiting one program and starting another.

NetWare 4.x
NetWare version 4.x was released in 1994 and offered a new centralized
administration service called NDS (Novell Directory Service). NDS not only
eliminated the need for the three separate programs of NetWare 3 (Syscon,
PCONSOLE, and FILER) but also allowed for administration of numerous
servers through one console. Prior to version 4, changes had to be made indi-
vidually on each server in the network. This was both a time-consuming and
cumbersome task.
The first release of NetWare 4.0 was fairly buggy and soon was replaced
with 4.1 and then 4.2, which was stable. Version 4.2 was released as a step-
pingstone toward version 5. It is worth noting that during the release of
version 4 Novell changed the name of the product to IntranetWare. Some
believe that this was a marketing ploy to take advantage of the Internet craze
that was forming at this time. The name was subsequently changed back
to NetWare in version 5.

NetWare 5.x
NetWare version 5 made a radical change in network communication for
Novell. Up to this point the Novell network protocol of choice was IPX/
SPX. With the release of version 5 Novell switched to TCP/IP as the protocol
of choice. This change was in part due to the sweeping support for TCP/IP
caused by the constant expanding Internet. The protocol of the Internet
is TCP/IP. NetWare 5 also included support for a multiprocessor kernel.
Previous versions of NetWare supported multiprocessors, but with the
addition of another NLM. Other added features included a five-version
license of Oracle 8 (a relational database software) and the inclusion of, which provided for management on the workstation side.

NetWare 6.x
A major improvement in the just released NetWare version 6 is the eDirectory,
an upgrade to the NDS structure introduced in Netware 4x. Although
eDirectory can be installed, maintained, and supported on non-NetWare
platforms, including Windows NT/2000 and Linux/Unix servers, NDS
runs only on NetWare servers. You no longer have to create and maintain

Copyright ©2002 SYBEX, Inc., Alameda, CA

Major Network Operating Systems 221

separate user accounts across platforms; instead, eDirectory automatically

updates user policies, profiles, and privileges across eDirectory-accessible
systems. This proves advantageous in multiserver environments where
having to update information across several different servers with different
operating systems can be problematic as well as time-consuming.
NetWare 6 also offers IPP (Internet Printing Protocol). As expected by its
name, this protocol allows for remote printing though the Internet.
Finally, version 6 introduces iFolders, which offers a connection between
locally stored data and server stored data that is synchronized. This allows
remote access to files along with data security implementation though
synchronous backup.
Because NetWare 6 was released just recently, there is still a lot of explo-
ration and learning happening. Within a year the good and bad will surface
and a clearer understanding of the power of the operating system will allow
for a clearer judgment on its effectiveness.

NetWare Architecture
All operating systems are modular to a degree. Key components make up a
core, which is the operating system. On this core other modules or programs
are added. Novell is a classic example of this design idea in practice. The
major component of NetWare is the kernel or core OS. Built on the kernel
are NLMs (NetWare Loadable Modules). By creating a structure such as
this, disk space can be conserved by selecting which components to load and
which to not load. There are four key NLMs: disk drivers, LAN drivers,
name space modules, and utility NLMs.
Disk drivers provide access to disks and disk resources. NetWare version 3
used a .DSK extension to allow access to IDE drives. With the release of
newer versions, this extension changed to .HAM or .CDM.
LAN drivers interface between the NetWare kernel and the network
card. This is an obviously important area, because the server must have a
connection to the network. These drivers typically have a .LAN extension.
Name space modules control how files look and are stored on the server.
By default NetWare stores files using the old DOS naming convention.
This is a file name of eight characters long, a period, and then a three-letter
extension. Because different client operating systems that are accessing the
server will use different file storage and naming conventions, the name space
modules act as a buffer to mediate between the different files. The extension
for a name space module is .NAM.

Copyright ©2002 SYBEX, Inc., Alameda, CA

222 Chapter 7  Network Operating Systems

Utility NLMs basically contain all the other items that do not fall into any
of the previously mentioned categories. More then 70 percent of NLMs
fall into the utility category, including print drivers (Novell Distributed
Print Services).
Each module is selectively loaded and linked to the NetWare kernel.
This creates the Netware operating system.

NetWare Administration
Administration for a NetWare server is actually done remotely. A separate
client utility called Novell’s NetWare Client for Windows 95/98 must be
installed on a client machine in order to control a NetWare server. This
allows the server to be physically locked up and secure. In NetWare version 3,
access is gained to the Bindery through a client machine running the Net-
Ware client software. You must then log into the server as an Admin or
a user with administrator rights. From this point you can gain access to cre-
ate users and manage the server. In NetWare versions 4, 5, and 6, use the
NDS rather than the Bindery. A Java applet can also be used rather then
the command line utility or menu-based monitor starting in version 5.

NetWare Interoperability
Novell NetWare supports Windows 95/98, Windows NT, Mac OS, VMS,
OS/400 Unix, and OS/2 clients. Prior to NetWare version 5, Novell ran
IPX/SPX natively. Starting with version 5, TCP/IP is the protocol of choice.

In comparison to the other major vendors, Microsoft is the new kid on the
block. And in being the new kid, Microsoft has taken a lot of heat over
the years due to the glitches and the number of repair software patches
that have been released to fix known problems.
Over the recent years Microsoft has released a multitude of operating
systems. The majority have been OSs for desktop computing but there have
been releases for network operating systems too. Windows NT was the first
Microsoft network operating system. It saw major changes and then was
replaced by Windows 2000. Repair of known issues as well as advanced
features and support for modern hardware are some of the key benefits
of Windows 2000.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Major Network Operating Systems 223

Major Versions of Windows

Microsoft originally focused their efforts on the desktop operating system
environment. However, wanting to expand into the growing network field,
they created a network operating system. Much like Novell’s NetWare, the
Microsoft product has seen rapid change and several versions released.

Windows NT
Microsoft first attempted to create a network operating system with the
release of Windows NT (New Technology). Version 3.1 was the first release
to the public and it came out in 1993. Much like the original release of Win-
dows 95, NT 3.1 was quite buggy and problematic. In the world of servers
that is a cardinal sin. For this reason NT 3.1 was really not taken very seri-
ously. NT 3.51 was released about a year later. Better stability, support for
new hardware, and a familiar interface (built with the Windows 3.1 GUI)
placed Microsoft into the network operating system arena that was previous
dominated by Novell and Unix. With time Microsoft released Windows 95
as well as NT 4, which was based around the Windows 95 GUI. The clever-
ness of Microsoft marketing shone through again. Using the GUI from an
operating system that the public was familiar with ensured less of a learning
curve in using NT 4 as well as eliminating some intimidation.
Stability issues and concerns from version 3 had been dealt with before
the release of Windows NT 4 in 1996, and NT 4 became increasingly pop-
ular. The gamble of creating a network operating system with a familiar
interface paid off and NT 4 vaulted Windows past Novell in network oper-
ating system sales as well as third-party programs being written for the OS.
Added components in version 4 included IIS (Internet Information Server), a
web server, and the Internet Explorer web browser. Windows NT 4 became
widely accepted as an OS for an enterprise server but not for a backbone
server—possibly due to fears dating from version 3’s instability, or possibly
the newness of this operating system compared to Novell or Unix.

Windows 2000
Windows 2000 Server was developed to address the fears in using Win-
dows as a true server operating system in large environments. With the
release of Windows 2000, a change in protocol use also occurred. Prior
to Windows 2000, Net BEUI was the native protocol of choice. Win-
dows 2000, much like NetWare, switched to TCP/IP as its main protocol
to be better supported with the Internet. Previous versions of Windows

Copyright ©2002 SYBEX, Inc., Alameda, CA

224 Chapter 7  Network Operating Systems

Server used NTDS (NT Directory Services) to control user accounts and
groups security. Windows 2000 switched to AD (Active Directory). This
new service was more in line with Novell and followed the X.500 standards,
using a hierarchical approach to naming conventions. Windows 2000 also
improved on network and system performance monitoring by creating a
management utility called Administrative Tools. Figure 7.1 is a screen shot
of Administrative Tools. Administration is now done centrally through the
MMC (Microsoft Management Console) in Windows 2000.

FIGURE 7.1 Windows 2000 Administrative Tools

Windows Interoperability
A server never operates in an isolated environment. Clients and other serv-
ers connecting to the server come from various sources and platforms such
as versions of Windows, NetWare, Unix, and Apple. Windows NT offers
client and print services for Novell, Apple, and Unix. The Novell services
are called GSNW (gateway services for NetWare), CSNW (client services
for NetWare), and FPNW (file and print services for NetWare). Macintosh
clients can also access a Windows server, but client software will have to be
installed. All Microsoft operating systems are also natively supported
(including DOS and Windows 3, 95, 98, and Me).

Copyright ©2002 SYBEX, Inc., Alameda, CA

Major Network Operating Systems 225

Windows Administration
Both NT and Windows 2000 are administered through the User Manager
for Domains. It is located in the Administrative Tools folder. Through this
utility you can create both users and groups for the local computer or for the
network. This simple utility gives a clear visual of the users on the network.

It is claimed that the first version of Unix was invented in 1969 at Bell Labs.
Regardless of the date, Unix is definitely the oldest network operating system
in use today. Even though there has been constant change in the appearance
and even function with each new flavor, the core of the operating system
remains the same.

Unix Architecture
Unix uses a 32-bit command-line-based core capable of supporting a GUI
(often called X Window). Unix natively supports TCP/IP as its primary
protocol. The Server+ Exam does not cover Unix operating systems very
much, but be aware of some of the essential flavors of Unix as well as
their benefits.

Major Versions of Unix

As previously mentioned, there are countless versions of Unix available.
The open design architecture combined with low cost of ownership has
encouraged the development of the many versions of Unix in active
use today.

Linux is a Unix flavor that has been receiving a lot of attention over the
last few years. Linux version 1 was released in 1994 and has been updated
constantly since that point. Linux prefers to be run on an Intel platform, but
successful attempts have been made with RISC processors as well as on a
Two releases of Linux should be mentioned: Red Hat Linux and Slack-
ware. Red Hat Linux is a portable version that will run on an Intel, Alpha,
and Sparc processor. Slackware was designed for the Intel platform only and
will support up to 16 processors, as well as Ethernet networking.

Copyright ©2002 SYBEX, Inc., Alameda, CA

226 Chapter 7  Network Operating Systems

For more information on Linux distributions go to

SCO Linux
Santa Cruz Operation Linux makes OpenServer and UnixWare. Open-
Server is robust and scalable and is used with Intel equipment. UnixWare
was obtained by SCO Linux from Novell in 1997. UnixWare provides
interoperability with Novell-based networks as well as being easy to
administer and install.

Sun Solaris
Sun Microsystems created its own Unix called Sun Solaris. It was designed to
run on a SPARC platform, rather than an Intel. Sun sells both the server
hardware and operating system together designed for Internet servers.

Unix Administration
Unix administration is most commonly done through a GUI called X Win-
dow or through a command line utility called a shell. There are three major
shells used: Bourne, C, and BASH. New accounts are made through modi-
fication of the /etc/passwd file. Normally administration is done through
the X Window utility, where the User Configurator is used.

Installing a NOS
I nstalling a network operating system can be a confusing process.
If you have ever formatted and reinstalled a desktop PC you will remember
that there are many processes and steps that have to be followed in a
specific order. Drivers and disks that support items such as video cards,
network cards, and SCSI cards will have to be installed with the assistance
of manufacturer-supplied drivers. At times, patch files and firmware will
have to be used as well. (Refer to Chapter 11, “Managing and Securing the
Server Environment,” for more information on patch files.)

Copyright ©2002 SYBEX, Inc., Alameda, CA

Installing a NOS 227

Verify Hardware Compatibility

The first step in installing a network operating system is to confirm that the
OS supports all the hardware. This will especially include the expansion card
devices such as video, SCSI, RAID, and network cards. Remember from
the previous section, some NOSs (such as Unix) are designed to run on
non-Intel-based machines. All connected hardware should be confirmed
as compatible before any attempt to perform an install is done. This also
includes external devices such as the UPS and UPS monitoring tools.
Verification at this stage will also include drivers for the software. Most
hardware will come with a CD or floppy disk containing drivers for common
operating systems. Oftentimes though the driver on the manufacturer disk
will be outdated—having been created at the time that the product was
made, these drivers can be several years old. It is always best to check with
the manufacturer’s website for the latest drivers. Other times the disk
may not contain drivers designed for the operating system that you are
installing. This also requires a trip to the manufacturer’s website to locate
new and suitable drivers. This is often the case today as people begin the
upgrade to Windows XP. The driver disks that came with their hardware
do not contain drivers for Windows XP. Although you can try to use an
NT or 98 driver, they probably will not work as expected.
Once all the proper drivers have been located and the hardware has been
confirmed with the operating system’s compatibility list, then the installa-
tion can occur. It is at this time that you will also have to deal with licensing
the server and determining the number of clients that are accessing it. Along
in the installation process the server will want to know how many client
licenses you have purchased. NOS installation is fairly straightforward:
NOSs are normally found on CDs that have an executable file that will begin
the installation process. Installation wizards then will guide you through the
installation process step-by-step. The Server+ Exam does not test you on
the details of each NOS install. Expect, with the multitude of configurations
(such as RAID), that the NOS installation will take a few hours to complete.
Once the network operating system has been installed, the system should
be rebooted (it will most likely want to reboot several times during the instal-
lation as well). This reboot will allow you to carefully monitor the computer’s
behavior. You should be looking for signs of trouble through the POST, error
messages, and beep codes.

Copyright ©2002 SYBEX, Inc., Alameda, CA

228 Chapter 7  Network Operating Systems

More on POST and error messages will be covered in Chapter 12, “Perfor-
mance and Hardware Monitoring,” and Chapter 15, “Disaster Recovery.”

Any problems during the installation of the operating system should be

noticeable and dealt with at this point. Windows 2000 offers the Computer
Management utility to help assess any hardware and software problems, and
Novell offers an ABEND.log to record any abnormal endings that the server
may have encountered. Both utilities point to potentially serious hardware
and operating system problems.

Verify Network Connectivity

Before installing any further software or network components, verification
of network connectivity should occur. Verification at this point takes on a
few forms. If you are using an Ethernet connection, there can be a visual
inspection of the link lights on the network card to ensure connectivity. Link
lights are located right beside the RJ-45 plug and they light up (usually green
or red) to indicate that there is connectivity between the network card and
the rest of the network. The next step is configuration of the server to the
network card and rest of the network. This will require naming the server.
The server should be provided with a meaningful name, as it is also the name
that is viewable to the clients on the network. For example, if the server is an
e-mail server, then the logical name should be Email Server. Other configu-
ration includes installing the appropriate domain name and protocol. If this
is a server for a brand new network, then the server might have to fulfill the
role of providing authentication and access to the network. In that sense (and
in accordance with Windows NT standards), it would be called a Domain
Controller. In selecting a protocol to control traffic and communication on
the network, a decision will have to be made as to the purpose of the network
as well as the type of connectivity. To keep things simple here (you will learn
more about protocols in Chapter 8, “TCP/IP”), it is best to use TCP/IP.

Configure External Peripherals

Now that the server is up and running, there are no hardware conflicts, and
the networking is communicating as expected, the time has come to install

Copyright ©2002 SYBEX, Inc., Alameda, CA

Installing a NOS 229

the external peripherals. This includes locally attached printers, modems,

external drives, and the UPS.
A locally attached printer can be used to print from a server or shared out
and used as a network printer. The latter case is commonly seen in small
networks. The server acts as a print server and a printer is locally attached
(or it can be attached through the network to the hub or switch). If you are
networking a printer off the server, the printer will have to be configured as
a network printer. This includes sharing the printer so it will be visible to
the clients, as well as installing a shared driver that will be accessible to the
clients who need to install the printer. Care must be taken to have available
drivers for all the client operating systems that might need the printer. For
example, a Novell server with Windows 98 and NT clients will need to
have printer drivers accessible for the Novell server, Windows 98, and NT
clients—three different server print drivers.
Installation of the UPS monitoring software and central connectivity
device monitoring tools occurs at this time. If you have a Cisco switch or
hub, there are remote management tools that can be installed on the server
to monitor and manage these devices.

Install Network Software

Network software includes items like database programs. Any software that
is installed on the server and will be used by clients is considered a network
software program. An example is an e-mail program. In in-house e-mail
programs, all messages are stored on the server until retrieved by the client.
This places a server in the role of a central storage facility requiring an
authentication method for the messages before releasing them.
After the previous step, confirming that the server hardware and
operating system are functioning properly and network connectivity is
running, all network-based software can be installed. Upon installing each
network program, confirm with the network operating system that the
software is operating successfully. This may be difficult without client
access, but that step can be verified at a later time.

Install NOS Updates to Design Specifications

Over time, updates are released by NOS manufacturers to repair known issues
or provide advanced features. These updates are provided by all manufacturers.

Copyright ©2002 SYBEX, Inc., Alameda, CA

230 Chapter 7  Network Operating Systems

It is a good idea to get on a mailing list with your OS manufacturer to keep

up to date. One of the key elements to remember is that not every update
produced must be installed on your server. Weigh the functionality and
need of the update against the potential risks and possible drawbacks and
see if it is really worth performing the update. Sometime it is best not to
take the chance. Other times, updates are essential. Windows 2000 was
not out very long before a service pack update was released to deal with
serious problems in the running of the operating system. This was an obvi-
ous patch that you would want to install. Windows NT, over the course
of its active existence, had six different service packs created and released
to the public to repair and improve on the operating system. Read the
information on the update carefully before you decide whether to install
or not.

For more information see Chapter 10, “Hardware Updates,” and Chapter 11,
“Software Updates.”

Client Access
Once you are sure that the operating system is running stably with the hard-
ware as well as network software, access can be granted to the clients slowly.
You may have wondered why, a few steps earlier, I mentioned installing net-
work software and partially testing it before there were clients accessing it to
ensure that it was running fully. Granting client access is usually the last step
when installing or upgrading a server. Until you are sure that the server is
fully functional and capable of handling the stress of client requests, you
should not allow access to it. Not only will you have to deal with an improp-
erly running server if you grant access too early, but everyone who has
connected to it will be calling you and letting you know that the server is
down. Clients may also take the liberty of starting to store files or use the
applications on the server, and if you need to restart the server for some
reason, you are going to get caught in the middle of some agitated personnel.
It is therefore best to ensure that everything is running ideally before granting

Copyright ©2002 SYBEX, Inc., Alameda, CA

Summary 231

Getting It All Running

Probably the best situation I have had the chance to work in was a few sum-
mers ago when the school I worked for decided to expand and purchase
a lot of new equipment. When the shipments arrived, I found myself on
the floor surrounded by five IBM Netfinity 5500 servers, five Intel 510T
switches, and 80 Dell laptops—all of which had to be networked. The nice
thing was that I had all summer to perform the task without anyone over my
shoulder asking if it was ready yet because they needed to check their
e-mail. Having the space and time planned out allowed me to carefully and
methodically unpack, inspect, install, and configure all the equipment. All
the hardware and software could be tested and run for days to confirm that
they would perform as expected. When problems did arise, there was
ample time to research and fix each situation.

Whether your project is a simple upgrade or a complete build, always leave

enough time to perform the task as well as any unexpected issues that
may arise.

C hapter 7 is all about network operating systems. Focus was given to
the major types of operating systems and their differences. We began with
an exploration of network operating system options including application
compatibility, hardware requirements, features, and cost. Each element was
discussed in comparison to the three major network operating systems:
Windows NT/2000, Novell NetWare, and Unix/Linux. OS/2 Warp, IBM’s
server operating system, was also introduced as another alternative.
Novell NetWare was released in 1983. It used a command-line-based
utility that was administered remotely from another computer. Major
releases of NetWare include version 3, which used a Bindery to maintain
usernames and accounts; version 4, which switched from the Bindery to
NDS (Novell Directory Services); version 5; and recently version 6. Novell
natively ran on IPX/SPX until version 5, which changed to TCP/IP.

Copyright ©2002 SYBEX, Inc., Alameda, CA

232 Chapter 7  Network Operating Systems

Windows released its server operating system called NT (New Technology)

in 1993. Version 3.1 was fairly buggy and not widely used. An update was
version 3.51, which repaired the issues in the previous version and used
the familiar Windows 3.1 GUI. NT version 4 updated numerous features,
including the use of the Windows 95 GUI. Windows 2000, released more
recently, included a new feature called the Active Directory to deal with users
and account security. Windows had natively run on the NetBEUI protocol
until Windows 2000, which changed to TCP/IP.
Unix comes in many flavors due to its open architecture and low cost of
ownership. Common releases include Linux, SCO Linux, and Sun Solaris.
Due to the uniqueness of each release, Unix tends to be immune to virus
attack. Natively, Unix runs on the TCP/IP protocol suite.
Steps involved in installing a network operating system include verifying
hardware compatibility. This includes compatibility with the server, external
devices such as modems and printers, and the rest of the network hardware
and software. Installing the network software includes programs such as
database and e-mail software that are used by the clients in the network.
Finally client access is given to the server after all hardware and software
functions are confirmed as operating properly.

Exam Essentials
Know the three main network operating systems. This includes Win-
dows NT/2000, Novell NetWare, and Unix.
Know the variables to consider when selecting a NOS. Application
compatibility, hardware requirements, features, and cost are the main
variables to consider.
Know the basic installation requirements for each major NOS. Novell
version 6 requires an Intel Pentium II or higher processor with 2GB of
hard drive space, 256 of RAM, a CD-ROM drive, and a VGA adapter.
Windows 2000 requires an Intel Pentium 133, 128MB of RAM,
1GB of hard drive space, a mouse, keyboard, CD-ROM drive, and
a network card.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Key Terms 233

Unix requirements vary but generally include an Intel-based processor,

64MB RAM, 500MB hard drive space, a CD-ROM, and a VGA adapter.
Know the major releases of Novell NetWare. NetWare version 3 used
a Bindery and was generally command-line-based.
Versions 4, 5, and 6 used NDS and offered a Java applet for administration.
Know how Windows is administered. Windows NT is administered
through User Manager for Domains, while Windows 2000 is adminis-
tered through Active Directory.
Know the major versions of Unix in use today. These include Linux,
SCO Linux, and Sun Solaris.
Know the steps to installing a server. Be able to verify hardware com-
patibility, verify network connectivity, configure external peripherals,
install network software, install network updates, and give client access
to server.

Key Terms
Before you take the exam, be certain you are familiar with the follow-
ing terms:

AD (Active Directory) GUI

BASH iFolders
Bindery IntranetWare
Bourne IPP (Internet Printing Protocol)
command line kernel
eDirectory Linux
flavors network operating system (NOS)

Copyright ©2002 SYBEX, Inc., Alameda, CA

234 Chapter 7  Network Operating Systems

NLMs (NetWare Loadable shell

Novell NetWare Slackware
NTDS (NT Directory Services) Sun Solaris
OpenServer Syscon
Per Seat UnixWare
Per Server Windows 2000
Red Hat Linux X Window

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 235

Review Questions
1. Which network operating system is referred to as a Warp Server?

A. NetWare
B. Unix/Linux

C. Windows 2000/NT
D. OS/2

2. Of all the network operating systems that we looked at in this chapter,

which one is the cheapest to acquire?
A. NetWare

B. Unix
C. Windows 2000/NT

D. OS/2

E. Linux

3. According to this chapter, Novell’s eDirectory can be used on which

operating systems?
A. NetWare

B. Unix/Linux

C. Windows NT
D. OS/2

E. All of the above

4. Which network operating system natively supports the Bindery

A. NetWare 3
B. Unix/Linux

C. Windows 2000

D. NT 4

Copyright ©2002 SYBEX, Inc., Alameda, CA

236 Chapter 7  Network Operating Systems

5. Which network operating system (or systems) is not administered with

a graphical user interface?
A. NetWare

B. Unix/Linux
C. Windows 2000/NT

D. OS/2

6. Which network operating system makes use of Active Directory

A. NetWare

B. Unix/Linux

C. Windows 2000
D. OS/2

7. Which network operating system has been used on e-mail servers

where stability, performance, and virus immunity were mandatory?
A. NetWare

B. Unix

C. Windows 2000/NT

D. OS/2

8. What is IPP an acronym for?

A. Internet Printing Protocol

B. Internet Program Passage

C. Intel Pilot Project

D. Interlaced Protection Program

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 237

9. Which of the following is not a Unix shell?

A. Bourne

B. Bash

C. X Window

10. What is the client utility that must be installed to administer a

NetWare server?
A. System 32


C. User Manager

D. Novell NetWare Client for Windows 95/98

11. Which version of Linux is designed for an Intel platform and will
support networking?
A. SCO Unix

B. Slackware

C. Red Hat

D. Susie

12. Which operating system uses a GUI in common with a desktop

operating system?
A. Windows NT

B. NetWare 5
C. OS/2 Warp

D. Linux

Copyright ©2002 SYBEX, Inc., Alameda, CA

238 Chapter 7  Network Operating Systems

13. What utility is used to administer the Bindery in NetWare version 3?

A. Syscon

B. Administrator

C. Bindery Manager

E. User Manager

14. Which operating system has been around the longest?

A. NetWare

B. Unix

C. Windows 2000/NT

D. OS/2

15. What is the minimum RAM requirement for Windows 2000?

A. 64MB

B. 128MB

C. 256MB

D. 32MB

16. Which NOS would you most likely find using Visual Basic scripts?

A. NetWare
B. Unix/Linux

C. Windows 2000/NT

D. OS/2

17. Which NLM would be used for network card drivers?

A. Name space
B. Utility

C. Disk


Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 239

18. Which of the following is not a NOS?

A. Windows 2000 Server

B. Windows NT 4

C. Novell NetWare
D. Windows Me

19. Which operating system uses NLMs?

A. Unix
B. Windows 2000

C. NetWare

D. OS/2 Warp

20. Which of the following network operating systems use TCP/IP as

their preferred transport protocol?
A. NetWare

B. Unix/Linux

C. Windows 2000/NT

D. OS/2

E. All of the above

Copyright ©2002 SYBEX, Inc., Alameda, CA

240 Chapter 7  Network Operating Systems

Answers to Review Questions

1. D. OS/2 from IBM is known as Warp Server.

2. E. Linux is available for the price of a download.

3. A, B, C. NDS eDirectory can be ported to Unix/Linux and Win-

dows NT, as well as run on its native NetWare.
4. A. NetWare 3 supports the use of the Bindery.

5. B. Unix and Linux are primarily administered through script or

configuration files.
6. C. Windows 2000 makes use of Active Directory Services.

7. B. As much as proponents of the other operating systems would love

to argue the point, Unix is the operating system of choice for most
mission-critical super-servers.
8. A. IPP is a program in NetWare version 6 that allows for Internet
based printing through the Internet Printing Protocol.
9. D. DOS is an operating system, not a shell of Unix.

10. D. Novell is administered through the use of a client computer with

the Novell NetWare Client for Windows 95/98 installed.
11. B. Slackware Linux was designed for the Intel platform and will
support Ethernet networking.
12. A. Windows NT is based around the Windows 95 GUI.

13. A. In NetWare version 3.X the Bindery was administered through

the Syscon utility.
14. B. Unix has been around for over 30 years.

15. B. Windows 2000 requires a minimum of 128MB of RAM to operate.

16. C. Visual Basic is a Microsoft product and is an integral part of the
Windows 2000 and Windows NT environment.
17. D. LAN NLMs are used to install network drivers for NetWare.
18. D. Windows Me is Microsoft’s home version operating system.

19. C. Novell’s NetWare is based around NetWare Loadable Modules.

20. E. All of the network operating systems mentioned in this chapter

would like your network to run on nothing but TCP/IP.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter TCP/IP



 6.2 Use diagnostic hardware and software tools and utilities

 Identify common diagnostic tools across the following OS:
Microsoft Windows NT/2000; Novell NetWare, Unix, Linux,
 Perform shut down across the following OS: Microsoft
Windows NT/2000, Novell NetWare, Unix, Linux, IBM OS/2
 Select the appropriate tool
 Use the selected tool effectively
 Replace defective hardware components as appropriate
 Identify defective FRUs and replace with correct part
 Interpret error logs, operating system errors, health logs,
and critical events
 Use documentation from previous technician successfully
 Locate and effectively use hot tips (e.g., fixes, OS updates,
E-support, web pages, CDs)
 Gather resources to get problem solved: Identify situations
requiring call for assistance; acquire appropriate
 Describe how to perform remote troubleshooting for a
 Describe how to perform remote troubleshooting for a
remote alert.

Copyright ©2002 SYBEX, Inc., Alameda, CA

N etworking protocols are one of the most important com-
ponents in a network environment, providing communication capabilities
between network computers. The most commonly used protocol on net-
works today is TCP/IP.
This chapter will introduce you to the fundamentals of the TCP/IP
protocol suite. It will then go on to explain the major concepts behind IP
addressing and introduce you to some of the utilities that can be used to
troubleshoot an IP-based network.

TCP/IP Explained
T CP/IP has is an Internet standard protocol that is implemented on
most networks today. It is a suite of different protocols that provide com-
munication capabilities between network computers. The following section
will provide you with a brief overview of the TCP/IP protocol suite and the
role each protocol plays in network communication.

What Is TCP/IP?
TCP/IP has become the most widely used protocol. It is the protocol used on
the Internet and on routed networks. TCP/IP is a suite of protocols that maps
to a four-layer reference model known as the DoD (Department of Defense)
Model. The Model basically describes how communication between two
hosts on a network occurs.
You may also be familiar with the OSI Model. This model was developed
to implement a standard on how network communication occurs. Vendors

Copyright ©2002 SYBEX, Inc., Alameda, CA

TCP/IP Explained 243

design their products based on the OSI Model. Essentially the OSI Model
enables communication between software and hardware regardless of the
vendor (as long as the network components are designed to the standards of
the model).
The OSI Model divides network communication into seven layers, as
opposed to the four layers used in the DoD Model. Each layer in the DoD
Model maps to one or more of the seven layers in the OSI Model.
The four layers of the DoD Model are as follows: Application, Transport,
Internet, and Network. The following section will discuss the different layers
of the DoD Model and how the different protocols making up TCP/IP map
to this four-layer model.

Application Layer
The top layer of the DoD Model is the application layer. This layer defines
how applications communicate with one another on a network. Application
layer protocols often provide support to client/server applications. A client
application running on one computer will communicate with the server appli-
cation running on another computer using the services of application layer
protocols. Let’s take a moment to review these protocols in detail.
Simple Mail Transfer Protocol SMTP (Simple Mail Transfer Protocol)
is used to send e-mail messages from one mail server to another. SMTP is
usually used to send e-mail messages over the Internet. The e-mail mes-
sages are then retrieved using POP or IMAP. When you configure your
e-mail application, you have to specify the SMTP server that your e-mail
application will be sending mail to.
Simple Network Management Protocol The Simple Network Manage-
ment Protocol is a set of protocols that are used for collecting information
about a network. SNMP agents are network devices such as computers,
routers, and bridges that gather information about themselves and return
the information to a system running a SNMP management program.
File Transfer Protocol The file transfer protocol is used to transfer files
between computer systems. It is a standard protocol that allows files to be
transferred between dissimilar systems. The user transferring files will
usually need a username and password with permission to access the file
unless a guest account is being used on the remote system. FTP uses the
TCP protocol to transfer the files.

Copyright ©2002 SYBEX, Inc., Alameda, CA

244 Chapter 8  TCP/IP

Trivial File Transfer Protocol TFTP is a simple version of the file

transfer protocol described above. It uses the UDP protocol and, unlike
FTP, provides no security. A username and password are not required
for TFTP.
Telnet Telnet is a terminal emulation protocol used on TCP/IP net-
works. A client computer can use Telnet to log onto and connect to
another host on the network regardless of the platform on which it is run-
ning. You can enter commands through Telnet and they will be executed
on the remote system just as if you were sitting at the console. Telnet is
often used for troubleshooting and connecting to and configuring routers.
Hypertext Transfer Protocol HTTP is used for communication between
a web browser and a web server. It is the protocol used on the Internet to
format and send information. HTTP also dictates what actions to take
when certain commands are issued through a web browser.

Transport Layer
The next layer of the DoD Model is the transport layer. The protocols at this
layer define the type of transmission service between two hosts. The trans-
mission services provided by this level can be end-to-end and reliable or
broadcast-based and unreliable. The protocols operating at this level provid-
ing transmission services include TCP and UDP.

Transmission Control Protocol

TCP (transmission control protocol) is a transport layer protocol that
provides guaranteed delivery of packets between two hosts. It provides
end-to-end communication, sequencing, and error control to ensure reliable
delivery of data.
Messages received from the application layer are broken down and
sequenced. By sequencing the packets, TCP on the destination host is able to
reassemble the message in the correct order and pass it to the upper applica-
tion layer. On the sending computer TCP sends a group of packets to the
destination host and waits for an acknowledgment for each one. Any packets
that have gone unacknowledged will be retransmitted.

User Datagram Protocol

UDP (user datagram protocol) is the other transmission protocol that works
at this layer. It is similar to TCP but it does not provide reliable delivery of

Copyright ©2002 SYBEX, Inc., Alameda, CA

TCP/IP Explained 245

data between hosts. It takes the messages from the application layer and
breaks them down into smaller segments and uses sequencing. The difference
is that UDP is not concerned with acknowledgments and does not resend
packets that do not reach the destination host. Reliable delivery and error
checking are performed by protocols working at the upper layers of the
DoD Model.

Internet Layer
Protocols operating at the Internet layer are responsible for things such as
IP addressing and routing. This layer adds addressing information to the
packet before it is placed on the network, resolves IP addresses to MAC
addresses, and determines where a packet will be sent in order to reach the
destination host. The protocols that work at this layer include IP, ARP,

Internet Protocol
IP is by far the most important protocol working at the Internet layer and
is often referred to as the mailroom of the TCP/IP protocol suite. It is here
where packets from the upper layer are broken down into datagrams. IP
adds addressing information to each datagram including the IP address of
the sending computer and the IP address of the destination computer (refer
to the section titled “IP Addressing” for more information). IP also performs
routing functions because it is responsible for determining the best route for
a datagram to be sent to reach the destination host.

Address Resolution Protocol

IP addresses are used for routing a datagram. Before a datagram can be sent
to a host on a network, the logical IP address must be resolved to the physical
address (or MAC address). If IP does not know the MAC address of the
recipient, it is the responsibility of ARP to resolve the IP address to the MAC
address. This is done via broadcast on the local subnet.
ARP maintains a local cache of IP addresses that have been resolved to
IP addresses. Before performing a broadcast, ARP will check the local cache
to see if the IP address has already been resolved. If not, ARP generates a
broadcast containing the IP address of the intended recipient. The intended
recipient may be the actual host the local computer wants to communicate
with or a local router if the destination host is on another subnet. All com-
puters on the local subnet will look at the broadcast, and the host with the
IP address will respond with its MAC address.

Copyright ©2002 SYBEX, Inc., Alameda, CA

246 Chapter 8  TCP/IP

The BootP (boot program) protocol allows diskless workstations to boot up
and send out a BootP request so the workstation can receive an IP address.
The broadcast contains the MAC address of the client. A BootP server picks
up the request and looks to the BootP file. If the BootP file contains an entry
for the workstation’s MAC address, the BootP server responds with the work-
station’s IP address and the name and location of the file it should boot from.

Internet Control Message Protocol

ICMP works with the IP protocol and is mostly a management and trouble-
shooting protocol. ICMP reports any packets containing error, control, or
information messages. It will notify the sending host of any abnormal events
that occur. The PING command, which is discussed later in the chapter, uses
this protocol to test the connectivity between hosts.

Internet Group Management Protocol

IGMP (Internet Group Management Protocol) is used for multicasting on
an IP network. It does this through the use of multicast groups. Hosts on a
network use the IGMP protocol to join or leave a multicast group (host

Network Layer
There are no protocols and services defined at the network layer but it does
perform some very important functions. Essentially, it takes datagrams from
the network layer and breaks them into bits (1’s and 0’s). Then it adds the
MAC address to the packet before it is placed on the network. Once placed
on the network, it will determine the access method used by the network (for
example, token passing, collision avoidance, collision detection). After it
defines the network’s access method, it shifts its focus to the physical aspects
of the network. The network layer determines the physical aspects of the net-
work, such as the media and connectors.

IP Addressing
N ow that you have an understanding of the protocols that make up
TCP/IP and what each one is responsible for, let’s move on to IP addressing.

Copyright ©2002 SYBEX, Inc., Alameda, CA

IP Addressing 247

There is a good chance that TCP/IP will be the protocol implemented on

your network. In order to effectively implement an IP based network, it is
important to have a basic understanding of IP addressing and subnetting.
Even if you are not implementing the actual addressing scheme, general
knowledge of how IP addressing is implemented can assist in troubleshoot-
ing addressing problems that may arise on the network.

Introduction to IP Addressing
IP addresses are used to identify hosts on a given network. Every host
(this includes computers, routers, and network interface printers) on an IP
network requires an IP address. Each segment on a given network requires
a unique network ID and every host on a segment requires a unique host ID.
An IP address can be seen in two different formats—as a decimal number
and as a binary number. Computers read an IP address as a 32-bit binary
number like the one shown below:
10110111 11101111 10110011 10101011
For computer users, the IP address is viewed in decimal format. The
binary number, like the one shown above, is broken down in four sections
(called octets), each section containing eight bits. Each of the four octets is
then converted to a decimal format by converting each bit to a number value
and adding them up, giving you an IP address like the one shown below:
Binary uses 1’s and 0’s. Each bit in a binary number has a corresponding dec-
imal value. The bit values within each octet are converted to a decimal format
and then totaled. Table 8.1 shows the decimal value for each bit in an octet.

TABLE 8.1 Converting from Binary to Decimal

Format Value Value Value Value Value Value Value Value

Binary 1 1 1 1 1 1 1 1

Decimal 128 64 32 16 8 4 2 1

Using this table, the binary number shown below can easily be converted
to the decimal number below it.
11000000 10101000 00011000 10000100

Copyright ©2002 SYBEX, Inc., Alameda, CA

248 Chapter 8  TCP/IP

Network and Host ID’s

An IP address consists of two parts, one identifying the network and
the other identifying a unique host on that network. The most common
analogy used to describe network and host ID’s is in terms of your own
personal address. Your address will consist of a street and house number.
All houses on the same street have the same street address and each house
on that street requires a unique house number. Similarly, all segments on
a given network require a unique network ID to identify them, and each
host requires a host ID that is unique to their segment. The next section
shows you how to determine which portion of an IP address represents the
network ID and the host ID.
For every IP address a corresponding subnet mask is required. The
subnet mask is used to distinguish which portion of the IP address refers
to the network ID and which portion of it refers to the host ID. When a
computer on a network needs to communicate with another computer, it
must determine if the destination host is on the same network or if it is on
a remote network.
The subnet mask is used to determine what part of the IP address is used
to identify the network ID. This is done by setting all the bits in the subnet
mask representing the network ID to 1’s (the remaining bits used for the
host ID are left as 0’s). For example:
IP Address 11000000 10101000 00001010 00000010
Subnet Mask 11111111 11111111 11111111 00000000
By looking at the subnet mask in the example above, you can see that the first
three octets of the IP address are used for the network ID (

IP Address Classes
Those who designed the Internet also broke the available IP addresses down
into different classes that could be assigned based on network size. The
number of hosts you can have per network address depends on the class
of address that you have been assigned. How you determine the network
and host ID’s within an IP address is also dependent on the address class.
Table 8.2 summarizes the five address classes.

Copyright ©2002 SYBEX, Inc., Alameda, CA

IP Addressing 249

TABLE 8.2 IP Address Classes

First Octet Max Hosts

Class Value Max Networks Per network Subnet Mask

A 1–126 127 16,777,216

B 128–191 16,384 65,534

C 192–223 2,097,152 254

D 224–239 Used for


E 240–254 Used for experi-

mental purposes

The value 127 is not included in the available ranges; 127 is reserved for test-
ing purposes and cannot be assigned to any computer. is referred to
as the loopback address and can be used to test whether TCP/IP is initialized
on a local computer by typing PING Essentially what the computer
is doing is pinging itself.

Certain IP address ranges from the different classes have been excluded
for use on the public Internet. These are known as private IP addresses and
are only used on private networks. If you implement a private IP address
range on your internal network but still want to access the Internet, you will
have to implement some form of gateway, such as Microsoft ISA Server, that
has an interface with a valid Internet IP address. Table 8.3 lists the private
IP address ranges.

TABLE 8.3 Private IP Address Ranges

Network Range Subnet Mask

Copyright ©2002 SYBEX, Inc., Alameda, CA

250 Chapter 8  TCP/IP

The problem with the current address classes is that they are very ineffi-
cient. For example, assigning a network of 10,000 users a Class A address
means that a large number of host addresses are going unused. A network
with 10,000 users being assigned a network address from a class that can
support up to 17 million hosts per network is inefficient. Also, with a typical
network today being essentially an internetwork—a single network made up
of multiple segments—how is routing going to occur if you have only been
assigned a single network ID. The following section will introduce the topic of
subnetting, which has been developed to overcome the current limitations
of the address classes.

Networks today are growing in size, spanning geographical locations, and
often consist of many different segments (networks within networks). You
may be asking yourself why you would want to divide a network into several
different distinct segments. Here are a few reasons:
 It allows you to connect different types of networks such as an
Ethernet network with a token ring network.
 Remote offices or locations can be made a part of the main network.
 Traffic can be limited to a local segment, thus reducing broadcast
If a network is to be divided into distinct segments, each segment will
require its own unique network ID. If you have a single network ID that has
been assigned to you by InterNIC, you must take that network ID and divide
it into other network ID’s.
Subnetting overcomes the limitation of a single network ID. Subnetting
involves taking a single network ID and splitting that into multiple subnets
(or multiple network ID’s). Each segment is separated by a router, and a
custom subnet mask is created so there is a way to distinguish between the
different subnets. Essentially what you are doing is taking a single network
ID and dividing it up into smaller subnets.
Since the subnet mask is used to determine the network ID of an IP
address, you will need to create a custom subnet mask. This is accomplished
by taking away some of the bits used for the host ID’s and creating a subnet
address. The bits that you are taking away from the host ID’s are now going
to be used to identify the different subnets within the network (this also
means that you will now have fewer host ID’s available on your network).

Copyright ©2002 SYBEX, Inc., Alameda, CA

IP Addressing 251

Take a look at a simple example to illustrate this: Suppose you have been
assigned a Class B address of The default subnet mask associ-
ated with a Class B address is Computer 1 and Computer 2
have been assigned the following IP addresses:
Computer 1 –
Computer 2 –
Using the subnet mask it can easily be determined that the two computers are
on the same subnet (by looking at the first two octets of each IP address to
see if they correspond). Now suppose the default subnet mask for the Class B
address is changed to the following because the network is divided into dis-
tinct segments and Computer1 and Computer2 reside on different subnets:
Computer 1 –
Computer 2 –
The custom subnet mask now indicates that these two computers are on
different subnets. Instead of looking at the first two octets of the IP address
(which is the default for a Class B address), the subnet masks now indicate
that the third octet is used to identify unique subnets. Since the decimal value
of the third octet does not correspond between the two IP addresses, the
computers are on different subnets.
Implementing a custom subnet mask may seem like a complex process but
it follows a very logical process. Use the steps outlined below to plan for your
subnetted network.
1. Determine the number of subnets or the number of network ID’s that
will be required taking into account future growth. A unique network
ID is required for each subnet and each WAN connection.
2. Determine the maximum number of host ID’s you are going to need
for each subnet. A host ID is required for each network card using
TCP/IP, any network interface printers running TCP/IP, and for each
router interface.
3. Create a custom subnet mask based on the above information that
provides the necessary number of subnets and hosts per subnet.
Let’s take a look at another example to illustrate the process of creating
a custom subnet mask. Suppose you have been assigned the network address
of and a subnet mask of (Class B). Normally the
first two octets would be used for the network ID and the last two octets for
the host ID’s.

Copyright ©2002 SYBEX, Inc., Alameda, CA

252 Chapter 8  TCP/IP

To create a custom subnet address to differentiate between the different

subnets on the network, you take away some of the host bits from the third
octet. The number of bits you take away for the subnet mask will depend on
the number of network ID’s you need. If you require six subnets, you may
end up with a custom subnet mask as shown below:
Binary 1111111 11111111 11100000 00000000

Notice that the binary value for the subnet mask has now increased by three
bits as compared to the default subnet mask.

Use the following steps to create a custom subnet mask:

1. Determine the number of network ID’s.

2. Convert the number of subnets (network ID’s) to binary format.

3. The number of bits represents the number of subnets required and this
is the number of bits that you need to add to the default subnet mask.
4. Once the required bits have been added to the subnet mask, convert
the number back to decimal value.
Let’s take a look at another example. Suppose you are assigned the Class
A address of and you need to create eight subnets. Your first step
is to convert 8 into binary format, which gives you 00001000. All bits to the
left of the leftmost one are ignored. The required number of bits then is 4 and
4 bits needs to be added to the default subnet mask, giving you a binary num-
ber shown below.
11111111 11110000 00000000 00000000
Convert this binary number back to a decimal format (255. 240.0.0)
to get your custom subnet mask that will be assigned to all hosts on each
distinct network.

IP Version 6
Since the Internet has grown in popularity, there is a shortage of IP
addresses. The discussion so far has been focusing on IP version 4. The
next version of IP, version 6 (IPv6), will overcome the addressing limitations
of version 4 by using a 128-bit address. It will no longer be expressed as a

Copyright ©2002 SYBEX, Inc., Alameda, CA

TCP/IP Utilities 253

four-octet decimal number, but rather as hexadecimal digits, with colons as

separators between the sets.

Assigning IP Addresses

Each host on a network requires an IP address. You can assign IP addresses

statically or dynamically. The option you choose will depend on the size of
your network. In a small environment, the administrative overhead of going
around to each computer to assign an IP address as well as any optional
parameters may be minimal. In a large environment, this may be impossi-
ble. In such cases you can install a service such as DCHP on one of your net-
work servers and give it a range of IP addresses that it can automatically
assign to clients. In terms of administration, it can all be done from a central
location; the IP address range and parameters need to be typed in just once,
and there is less chance for human error.

When assigning IP addresses, network servers should always be config-

ured with the same IP address. By creating a client reservation on the
server assigning IP addresses, your network servers can still have dynamic
IP addresses—but with a client reservation, they will always be assigned
the same one. On the other hand, having your servers dependent on
another server for their IP addresses does create a single point of failure.
If the server assigning IP addresses is unavailable, your servers may not
be able to acquire an IP address.

TCP/IP Utilities
There will be instances in a network environment when you experi-
ence TCP/IP connectivity problems, such as a client is unable to see a server
on the network. TCP/IP comes with several utilities that can be used to trou-
bleshoot connectivity problems that do arise. You should be familiar with
these tools if you administer an IP-based network or if your internal network
is connected to the public Internet.

Copyright ©2002 SYBEX, Inc., Alameda, CA

254 Chapter 8  TCP/IP

Some of the common TCP/IP utilities that are used for troubleshooting
include the following:
Many of these utilities can perform various functions depending on the
switches you use with the command. The switches available may depend on
the operating system you are running. The following section will outline
each of these utilities and how they may be useful in troubleshooting com-
munication in a server environment.

ARP stands for address resolution protocol. When a host on an IP-based net-
work wants to send data to another host, the host name must be mapped to
an IP addresses and the IP address mapped to a MAC address. ARP is the
protocol responsible for mapping IP addresses to MAC addresses. It does
this by sending out a broadcast packet containing the IP address of the
intended host. The host owning the IP address responds to the broadcast
with its MAC address.
Each host also maintains an ARP cache containing IP-to-MAC address
translations. Before a broadcast is made to resolve a MAC address, the ARP
cache is examined to see if there is an entry. The main purpose of this is to
cut down on the amount of broadcast traffic on a network. The syntax for
viewing the ARP cache on a local computer is as follows:
Arp -a

Copyright ©2002 SYBEX, Inc., Alameda, CA

TCP/IP Utilities 255

By default, entries stored in the ARP cache are dynamic, meaning they are
not permanent and are deleted within two minutes if not referenced. Using
the -s command you can add in a static entry that will not be flushed out.
This is useful for frequently accessed hosts. When adding the entry you must
include the IP address of the host as well as the MAC address. Once the entry
has been added it is listed as static. Table 8.4 summarizes some of the other
switches that can be used with the ARP command.

TABLE 8.4 ARP Command Line Switches

Switch Function

-a Displays the current entries in the ARP cache

-s Adds a static entry to the ARP cache for the Internet address
and physical address specified

-d Deletes an entry in the ARP cache

-n Displays the ARP entries for the specified Internet address

One of the most commonly used utilities when troubleshooting connectivity
problems on an IP-based network is the PING (Packet Internet Groper) util-
ity. It is a command line utility that is primarily used to see if another host
on the network is reachable and responsive. It works by sending out packets
to another host on the network and waits for a reply.
Most TCP/IP stacks that come with the different network operating sys-
tems come with a PING utility. They function the same, with some small
variations such as the confirmation message that is received.
The general syntax of the PING utility is as follows:
Ping w.x.y.z
where w.x.y.z is the IP address assigned to the host that you are testing
connectivity with. If the host is reachable and responding, you will receive a
confirmation message (see Figure 8.1).

Copyright ©2002 SYBEX, Inc., Alameda, CA

256 Chapter 8  TCP/IP

FIGURE 8.1 PING statistics indicating no packet loss

If the host is unreachable or is not responding, you will receive a request timed
out message or a destination host unreachable message (see Figure 8.2).

FIGURE 8.2 PING statistics indicating request timed out

Once TCP/IP is installed and configured on a server (or client), you can
use the PING utility to determine if some of the parameters, such as the
default gateway, are correctly configured. If your network is subnetted and

Copyright ©2002 SYBEX, Inc., Alameda, CA

TCP/IP Utilities 257

connected via multiple routers, you can use this utility to ensure that remote
networks are reachable. For example, if your network configuration is sim-
ilar to the diagram shown in Figure 8.3, you can use PING to make sure that
TCP/IP is correctly configured on Server1 and that your server can commu-
nicate with servers on remote subnets using the steps outlined below.

FIGURE 8.3 Sample network configuration with two routers

Server 2 Router 2
Router 1

Server 1

1. Open the command prompt on Server1 and PING the IP address of

Server2 (ping If you get a reply message, then TCP/IP
and its parameters are correctly configured. If the PING to Server2
fails, you can continue using PING to determine where the connectiv-
ity problem is.
2. Next, PING the IP address of Server2’s gateway. If you receive a reply,
you can determine that there is a connectivity problem with Server2.
If you are unsuccessful, proceed to the next step.
3. From Server2 PING the IP address of the remote gateway (ping If you receive a reply, you can then determine that
Router 2 is where the connectivity problem lies. If the PING fails, you
need to determine which side of the router is not functioning.
4. PING the IP address of Server1’s default gateway. If you are unsuc-
cessful, Router 1 may be experiencing problems, but also check that
the TCP/IP parameters (such as the subnet mask and default gateway)
are correctly configured on Server1.

Copyright ©2002 SYBEX, Inc., Alameda, CA

258 Chapter 8  TCP/IP

Tracert is another TCP/IP command line utility that can be used to trace
the router interfaces that a packet must pass through to reach a destination.
The number of routers that a packet must pass through is displayed as a
hop count. Each router that forwards the packet is one hop. Tracert can be
useful for troubleshooting if you are unable to reach a destination host—for
example, if a router between your server and the destination host is having
problems. Using the utility will allow you to easily determine the router on
the network that is failing to forward packets (see Figure 8.4). Tracert can
also report the time it takes for a packet to reach its destination, which
can be useful in determining how efficient a specific route is.

FIGURE 8.4 Tracert helps you pinpoint traffic problems

Ipconfig is a command line utility that can be used to view the TCP/IP
parameters assigned to a host, such as the IP address, subnet mask, and
nameservers. Ipconfig is very useful on Microsoft operating systems but will
not work on NetWare or Unix systems. To use the utility, type ipconfig
from a command prompt. To view more detailed TCP/IP parameters
assigned to a host, use the /all switch along with the command. This com-
mand can be particularly useful if dynamic addressing is being used. For
example, you can use the command to view and troubleshoot the parameters
that have been assigned.

Copyright ©2002 SYBEX, Inc., Alameda, CA

TCP/IP Utilities 259

Netstat can be used to view the current TCP/IP inbound and outbound
connections on a computer. It can also provide you with information on
listening ports, TCP or UDP connection stats, Ethernet stats, and it can
display the contents of the routing table. Figure 8.5 shows the output of the
Netstat command using the -n switch. Table 8.5 lists the switches available.

FIGURE 8.5 Output of Netstat command

TABLE 8.5 Netstat Command Line Switches

Switch Function

-a Displays all connections and listening ports

-e Displays Ethernet statistics

-n Displays addresses and port numbers in numerical form

-r Displays the routing table

-s Displays per protocol statistics

Copyright ©2002 SYBEX, Inc., Alameda, CA

260 Chapter 8  TCP/IP

Telnet is a remote terminal emulation program that is most often used for
troubleshooting on TCP/IP networks. The Telnet program runs on a work-
station and is used to connect to another host such as a network server. Once
you establish a session with a host using Telnet, you can enter commands
within the Telnet console that will be executed on the remote host or server.
Telnet is often used in Unix environments as well as for configuring and
troubleshooting devices such as routers and switches.

The Nbtstat utility is used to view NETBIOS over TCP/IP statistics, such
as any current connections using NETBIOS over TCP/IP, protocol statistics,
and any NETBIOS names that have been resolved to IP addresses. Table 8.6
summarizes the switches available with this command line utility.

TABLE 8.6 Nbtstat Command Line Switches

Switch Function

-c Displays the local NETBIOS cache

-n Lists local NETBIOS names

-r Lists the NETBIOS names resolved via broadcast and WINS

-R Purges and reloads the cache

In this chapter you learned about the fundamental concepts underlying
the TCP/IP suite of protocols. Since TCP/IP plays such an important role in
networking and server environments, it is important to have some under-
standing of it. TCP/IP maps to the four-layer DoD Model and provides com-
munication services between hosts on a network. Each of the protocols
included in the suite operates at one of the four levels performing a specific

Copyright ©2002 SYBEX, Inc., Alameda, CA

Exam Essentials 261

The chapter also covered the concepts of IP addressing and subnetting.

Even if you are not setting up the addressing scheme, it is important to under-
stand how IP addressing works for troubleshooting purposes. All hosts on
a network require an IP address and a subnet mask. The IP address is made
up of a network ID and a host ID. The subnet mask is used to determine
whether two hosts are on the same network. To overcome some of the
limitations of IP addresses, you can implement subnetting.
Finally we looked at some of the utilities that can be used on an IP-based
network. There will be times when you experience communication problems
on an IP network. The utilities discussed can be useful in troubleshooting
and determining the source of the problem.

Exam Essentials
Understand the DoD Model. TCP/IP maps to a four-layer conceptual
model known as the DoD (Department of Defense) Model. It defines
four layers, each layer performing specific functions enabling network
Understand the TCP/IP protocols. TCP/IP is a suite of protocols. Each
of the protocols has a specific role and operates at one of the three upper
levels of the DoD Model. Each protocol plays a role in network commu-
nication between two hosts.
Know the difference between TCP and UDP. TCP and UDP work at
the transport layer providing transmission services. TCP is an end-to-end
connection-based protocol that provides reliable delivery of information
between hosts. UDP is broadcast based and does not provide reliable
delivery of information.
Understand IP addressing and subnetting. Every host on an IP network
requires an IP address and a subnet mask. This is a 32-bit logical number.
It is made up of a network ID and a host ID. To overcome the limitation
of IP addressing, you can implement subnetting. This allows you to take
host bits away for a subnet ID.
Understand the function of the different TCP/IP utilities. TCP/IP
supports many different utilities that can be used in troubleshooting
an IP network.

Copyright ©2002 SYBEX, Inc., Alameda, CA

262 Chapter 8  TCP/IP

Key Terms
Before you take the exam, be certain you are familiar with the follow-
ing terms:

default gateway SMTP (Simple Mail Transfer Protocol)
IMAP subnet mask
IP TCP (transmission control protocol)
Ipconfig TCP/IP
Nbtstat Telnet
Netstat UDP (user datagram protocol)

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 263

Review Questions
1. Which of the following TCP/IP protocols provide end-to-end reliable
communication between hosts?



2. Which of the following utilities can be used to view the TCP/IP con-
figuration parameters on a Windows based computer?
B. Ipconfig

C. Netstat

D. Nbtstat

3. What TCP/IP utility will display a list of all routers a packet must pass
through to reach a destination host?
A. Telnet



D. Tracert

4. What is TCP/IP?
A. A protocol used in network communication

B. A troubleshooting utility

C. A datagram sent from one host to another

D. A suite of protocols for network communication

Copyright ©2002 SYBEX, Inc., Alameda, CA

264 Chapter 8  TCP/IP

5. You have just installed TCP/IP on one of your network servers. You
want to test to make sure that it is initialized on the server. What
command can you use?
A. Ping
B. Tracert hostname

C. PING hostname

D. Ipconfig

6. What Internet layer protocol is used by the PING utility?




7. What TCP/IP protocol is used to resolve logical addresses into

physical addresses?




8. What is the default subnet mask for a Class B IP address?




9. You are a network administrator for a medium-size company. One of

your responsibilities is to administer mail servers. One of the depart-
ments reports it is unable to receive and send e-mail. At which layer in
the DoD Model is the problem occurring?

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 265

A. Network

B. Internet

C. Transport

D. Application

10. Your computer is assigned the IP address Which of

the following represents this number in binary format?
A. 10000011 01101011 00110110 11011101
B. 10000011 01110111 00110110 11011101

C. 10000001 01101011 00110110 11011101

D. 01000001 10010001 01110111 10001010

11. Which of the following utilities can be used for remote

A. Nbtstat

B. Telnet

C. Netstat

D. Tracert

12. What is the default subnet mask assigned to an IP address of


13. What DoD layer is responsible for taking datagrams and breaking
them up into bits, before adding the MAC address and placing the
data on the network media?
A. Network
B. Application

C. Internet

D. Transport

Copyright ©2002 SYBEX, Inc., Alameda, CA

266 Chapter 8  TCP/IP

14. Which protocol performs the same function as TCP except it is




15. Which of the following protocols is responsible for message address-

ing and routing functions?



16. By default, what is the host ID for the IP address of





17. By default, what is the network ID for the IP address of



18. What utility allows you to display all the current connections for a

B. Netstat
C. Nbtstat


Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 267

19. Which decimal value represents the binary number 11100000

11111110 10101011 00011010?
A. 224. 254.171.26



20. Which of the following TCP/IP utilities will allow you to view statis-
tics on NETBIOS over TCP/IP?
A. Netstat

B. NetBIOSstat

C. Nbtstat
D. Nbstat

Copyright ©2002 SYBEX, Inc., Alameda, CA

268 Chapter 8  TCP/IP

Answers to Review Questions

1. C. The Transmission Control Protocol provides end-to-end reliable
delivery of packets between hosts through session establishment,
sequencing, and error control.
2. B. The Ipconfig utility can be used to see the IP configuration param-
eters that have been assigned to a computer, such as the IP address,
subnet mask, and default gateway.
3. D. The Tracert utility can be used from a command prompt to trace
the router interfaces a packet must pass through in order to reach a
4. D. TCP/IP is a suite of protocols for network communication. It is
made up of several protocols that perform specific functions enabling
network communication.
5. A. PING is used to test if TCP/IP is initialized on a local
6. C. The Internet Control Message Protocol is used by the PING utility.

7. B. The address resolution protocol is used to map logical addresses

into physical addresses.
8. B. The default subnet mask for a Class B network is

9. D. SMTP and POP are responsible for the sending and receiving of
e-mail messages. Therefore the problem would lie at the application
10. A. When is converted to binary, the correct binary
number is that of answer A.
11. B. Telnet will allow you to connect to a remote device such as a server
or router and perform remote administration tasks.
12. B. is a Class B address. Therefore the subnet mask
will be
13. A. Datagrams are broken up into 0’s and 1’s and addressed at the
Network layer of the DoD model. The data is correctly addressed with
the MAC address before being sent out on the media to its destination.
14. A. UDP performs the same function as TCP only it is a connectionless

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Review Questions 269

15. B. IP is the protocol responsible for addressing messages and routing

16. D. The host ID for the Class C address of is
The first three octets are used for the network ID.
17. D. The network ID for the Class B address of is The first two octets identify the network and the remain-
ing two identify hosts.
18. B. The Netstat utility will allow you to view all current connections
for a server (or workstation).
19. A. The binary number represented in decimal format is
20. C. The Nbtstat utility can be used to view statistics on NETBIOS
over TCP/IP.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Upgrading PART


Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Upgrading and

 1.2 Install hardware using ESD best practices (boards, drives,

processors, memory, internal cables, etc.)
 Mount the rack installation
 Cut and crimp network cabling
 Install UPS
 Verify SCSI ID configuration and termination
 Install external devices (e.g., keyboards, monitors,
subsystems, modem rack, etc.)
 Verify power-on via power-on sequence

 2.6 Update manufacturer specific drivers

 2.8 Perform Server baseline

 3.3 Add hard drives
 Verify that drives are the appropriate type
 Confirm termination and cabling
 For ATA/IDE drives, confirm cabling, master/slave and
potential cross-brand compatibility
 Upgrade mass storage
 Add drives to array
 Replace existing drives
 Integrate into storage solution and make it available to the
operating system

Copyright ©2002 SYBEX, Inc., Alameda, CA

 Perform upgrade checklist, including: locate and obtain
latest test drivers, OS updates, software, etc.; Review FAQs,
instructions, facts and issues; test and pilot; schedule
downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and baseline;
document the upgrade.

 3.5 Upgrade BIOS/firmware

 Perform upgrade checklist including: locate and obtain
latest test drivers, OS updates, software, etc.; review FAQs,
instructions, facts and issues; test and pilot; schedule
downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and baseline;
document the upgrade

 3.6 Upgrade adapters (e.g., NICs, SCSI cards, RAID, etc.)

 Perform upgrade checklist including: locate and obtain
latest test drivers, OS updates, software, etc.; review FAQs,
instructions, facts and issues; test and pilot; schedule
downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and baseline;
document the upgrade

 3.7 Upgrade peripheral devices, internal and external

 Verify appropriate system resources (e.g., expansion slots,
IRQ, DMA, etc.)
 Perform upgrade checklist including: locate and obtain
latest test drivers, OS updates, software, etc.; review FAQs,
instructions, facts and issues; test and pilot; schedule
downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and baseline;
document the upgrade

 3.8 Upgrade system monitoring agents

 Perform upgrade checklist including: locate and obtain
latest test drivers, OS updates, software, etc.; review FAQs,
instructions, facts and issues; test and pilot; schedule

Copyright ©2002 SYBEX, Inc., Alameda, CA

downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and baseline;
document the upgrade

 3.9 Upgrade service tools (e.g., diagnostic tools, EISA

configuration, diagnostic partition, SSU, etc.)
 Perform upgrade checklist including: locate and obtain
latest test drivers, OS updates, software, etc.; review FAQs,
instructions, facts and issues; test and pilot; schedule
downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and baseline;
document the upgrade
 3.10 Upgrade UPS
 Perform upgrade checklist including: locate and obtain
latest test drivers, OS updates, software, etc.; review FAQs,
instructions, facts and issues; test and pilot; schedule
downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and baseline;
document the upgrade

 4.1 Perform regular backup

 4.2 Create baseline and compare performance

 4.5 Perform hardware verification

 4.6 Establish remote notification

Copyright ©2002 SYBEX, Inc., Alameda, CA

I n a network environment, servers play one of the most important
roles. If a server is down or if it is performing poorly, it has a direct impact
on the rest of the network. Network servers are expected to be up, running,
and meeting the demands placed on them almost 100 percent of the time.
With these points in mind, your servers will need to be upgraded at some
point in the future, whether it is a software upgrade, a hardware upgrade, or
both, to continue to meet the demands being placed on them.
Once a server is installed and running, it can’t be forgotten. All too often
servers are set up and then ignored under the motto, “If it is running fine,
then leave it alone.” Much like a car, servers need to be maintained. This
maintenance includes monitoring, assessment, and—over time—upgrading.
Through this process, hardware, software, and operating systems will be
upgraded, changed, and even retired as they age. If this process is carried out
properly, then unpredictable changes and financial strain can be avoided.
This chapter explores the process of assessing and maintaining a server
once it is up and running. This process can be carried out through several
means, including establishing baselines and reviewing monitoring agents
and log files. Based on gathered information, server performance can be
monitored. Over time with added client stress the need for an upgrade can
be determined, through this monitoring, before server performance degen-
eration results in either client frustration or server failure. The following
chapter will also introduce you to the procedures that should be followed
any time you are performing an upgrade to a server. Reviewing the infor-
mation discussed in the chapter can help to ensure that your upgrade is
successful and that you are prepared for any glitches that can occur.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment 277

One of the first steps in beginning to upgrade a server is to determine
exactly what needs to be upgraded. This of course will depend on the role the
server plays on the network and your reasons for performing the upgrade.
If a server has become a bottleneck, you will need to determine what server
component needs to be upgraded. Not performing an assessment can lead
to an unnecessary upgrade being performed and money being misspent. The
following section will look at performing an assessment of your server to
determine when to upgrade and how to determine what component to
Assessment should not be a one-time process; instead, the information
gleaned from constant monitoring can be useful for planning preventative
maintenance and determining upgrade paths.
Monitoring tools provide a means of watching and assessing server
performance. Windows 2000, for example, offers a program called System
Monitor, which is located in the Administrative Tools folder within the
Control Panel. Figure 9.1 is a screen shot from the System Monitor.

FIGURE 9.1 Windows 2000 System Monitor

Copyright ©2002 SYBEX, Inc., Alameda, CA

278 Chapter 9  Upgrading and Maintenance

Notice how the System Monitor offers a visual graph to represent current
performance of the computer. The information represented in the graph is
user-selected. In the screen shot example, processor time is the option
selected. To select an option simply click the plus button and choose the
counter you want to monitor. Figure 9.2 illustrates the Add Counters screen.

FIGURE 9.2 Add Counters screen in Windows 2000 System Monitor

With the System Monitor numerous counters can be used at one time.
This allows data to be collected and compared on several areas at once. The
data accumulated can then be analyzed to locate patterns in system behavior.
For example, does the processor time increase affect the interrupts-per-
second counter?
Besides the System Monitor, the Windows 2000 Performance utility also
shows Performance Logs and Alerts, including Counter Logs, Trace Logs,
and Alerts.
The Counter Logs Monitor allows you to set a period of time during
which the information gathered in the System Monitor will be recorded in a
text file on the hard drive. This allows you to maintain a record of perfor-
mance. This record can be reviewed at a later date or used to compare to data
gathered at another time.
The Trace Logs Monitor lets you trace changes and actions occurring on
a computer. This information is useful in locating the source of potential
problems. If a program action is responsible for server stress, then the Trace

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment 279

Log will assist in tracking the program through its utilization of server
The Alerts utility provides a means of setting a threshold on counters.
For example, the processor utilization seen in Figure 9.1 can have an alert set
so an action is performed should the number exceed a user-set value. This
action can include logging a message, sending a network message, starting
a performance data log, or running a program. With an alert set, server
stress can be properly handled before it leads to a bigger problem such as
system failure.
Regardless of the monitoring tools that are used (numerous third-party
software packages are also available), the idea is to seek out potential bot-
tlenecks. Bottlenecks are locations where the performance is hindered due
to poor performance. This can include processor speed, amount of RAM,
and hard drive speeds. System monitoring will assist in locating these prob-
lem areas. Monitoring tools used over a period of time will also assist in
locating areas in which performance is slowly decreasing. This is a common
problem that often goes unnoticed until it has become a serious issue.
Assessment also takes on the form of analysis of both the need to upgrade
as well as the upgrade procedure itself. With the assistance of Performance
Logs and Alerts, the decision to upgrade computer hardware or software can
be planned and carried out before poor server performance becomes critical.
This saves you the hassle of dealing with an emergency upgrade at a less than
opportune time. Since most servers are extremely expensive and complex
devices, it requires careful planning and assessment before deciding on an

Why Upgrade
As time goes on, networks change, and changing networks have a direct
impact on servers. Whether it is an increase in size or the use of a new net-
work service, these changes have a direct impact on the workload placed on
the server. To meet the demands of a changing environment and to be able
to perform under the expected workload, servers will need to be upgraded at
some point. This upgrade may be a hardware upgrade, a software upgrade,
or both.
You may think that the only reason you would upgrade a server is because
a component is failing or becoming a bottleneck, but there are several other
reasons as well. The following topics describe a few of the reasons why you
might need to upgrade a server.

Copyright ©2002 SYBEX, Inc., Alameda, CA

280 Chapter 9  Upgrading and Maintenance

Server Role
As you saw in Chapter 1, a server can play many different roles in a network
environment. The type of a role that a server is playing will determine its
hardware and software requirements. For example, one of the most impor-
tant components in a file server is the disk subsystem, but the most important
component in an application server is the processor. Changing a server from
one role to another may require that the server be upgraded.

New Applications and Services

Introducing new applications and services into a network environment can
mean that the servers on which they will be running need to be upgraded.
Most applications and services come with a list of hardware and software
requirements. Meeting the requirements might mean upgrading a server.
For example, if you plan to install Microsoft Cluster Service on a server,
you might need to first upgrade the operating system (to Windows NT 4
Enterprise Edition or Windows 2000 Advanced Server) and add the latest
service packs.
The same holds true for Novell and Unix applications as well. Always
make sure that the network operating system supports the software that you
are trying to install. Also, it’s critical to make sure the hardware you have
supports the software you are trying to add. Generally speaking, the stated
minimum hardware requirements are not sufficient. Always try to have more
hardware than the application or service requires, for better performance.

Network Growth
An increase in the size of a network will directly impact the workload being
placed on your servers. An increase in the number of users means an increase
in the number of users accessing network servers, whether it is for logon
validation, accessing shared resources and applications, or other services.
A server that was once capable of handling the workload might now need to
be upgraded to meet this increase in demand.

Hard Disk Space

One of the most common components to be upgraded in a server is the hard
disk. For example, if your server is acting as a file server, hosting shares and

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment 281

storing user data, you may quickly run out of available storage space and
find yourself needing to upgrade.

What to Upgrade
The examples given above describe just a few of the possible reasons for
upgrading hardware components and software on a server. Regardless of the
type of upgrade you are performing, preplanning is a must.
Determining what server component/components need to be upgraded
can be a difficult task and will depend on your reason for performing the
upgrade. If you are changing the role the server plays, you will need to assess
what components are most important for that given role. If you plan to make
an existing server a file server, you will probably be most concerned with
upgrading the server’s disk subsystem. If your reason for upgrading is poor
server performance, you will need to take some time assessing the server to
determine which component is causing the bottleneck. This is where a server
baseline becomes important (which also falls under the category of server
Server baselines can be used by network administrators to gauge the per-
formance of a server over time. The baseline helps to establish what normal
or acceptable performance for a server is. Using the information gathered
over time, you can monitor the different server components to see if they are
functioning individually and as a whole.
You’re not going to spend all of your time establishing baselines. You will
want to establish one when you first get the server running properly, and
then each time you add a component or make a major change, you will want
to re-establish new baselines as necessary. This will allow you to see how the
performances of different components change over time (establishing server
baselines will be covered in much more detail in Chapter 12, “Performance
and Hardware Monitoring”). Once you see deterioration in server perfor-
mance, you can monitor your server, compare it to your baseline of accept-
able performance, and use the comparison to determine what component (or
components) has become a bottleneck and what may need to be upgraded.
Once you’ve determined what needs to be upgraded within a server, you
can then begin planning for the actual upgrade. The tasks that need to be
completed before doing the upgrade will be covered in the following section.

Copyright ©2002 SYBEX, Inc., Alameda, CA

282 Chapter 9  Upgrading and Maintenance

What to Upgrade

Monitoring the real-time performance of a server on a regular basis can give

you a really good idea of how your server performs throughout the day
under various workloads. The information you gather and log can be valu-
able when trying to determine the source of a bottleneck and then what
server component needs to be upgraded. Be careful though when you are
using this information to determine the source of a problem. Upgrading one
server component that you think is causing a bottleneck can inadvertently
cause another bottleneck. When determining the source of a bottleneck, it is
important to take a holistic approach. You can analyze the performance of
server components individually but it is also important to analyze the server
as a whole and how the different components work together. For example,
suppose you run a performance-monitoring tool and determine that you
have a high number of I/Os. So you think the hard drive is the bottleneck and
upgrade it. Once the upgrade is complete, the number of I/Os does not
decrease and you determine through further analysis that there is actually
insufficient amount of RAM in the server. Not only have you spent dollars
on an unnecessary hard drive—now you have to spend more to upgrade
the RAM.

Upgrade Procedures
As previously mentioned, carrying out an upgrade on a server is not some-
thing to be decided on lightheartedly. A careful plan of the upgrade, includ-
ing a timeline, due dates, and milestones, must be established. During this
planning stage, consideration of compatibility between the upgrade and
existing hardware, software, and operating system must be dealt with, as
well as the clients who access the server. Will their hardware, software, and
operating systems be compatible with the server’s upgrade?
In order to properly and successfully perform a server upgrade, it is
important to follow some sort of procedural checklist. These are just some
basic but important steps that should be followed any time an upgrade
procedure is performed on a server, regardless of whether it is a hardware
upgrade or software upgrade. Preplanning may seem like a tedious task to do
for an upgrade but, when dealing with network servers, it is always better to
be safe than sorry. Take for example an operating system upgrade. This
is one of the most important upgrades that can be performed on a server

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment 283

and considerable preplanning is required. You may think the upgrade is

as simple as putting the CD in the CD-ROM, but what happens if the
upgrade fails? What about the availability of the server’s resources during
the upgrade? What about the applications installed on the server? Will the
server’s hardware be compatible with the new operating system? Preplan-
ning and completing the tasks outlined below will provide you with the
answers to these kinds of questions before the upgrade takes place.

Upgrade Nightmare

I remember a few years ago being called out to an office where an operating
system upgrade was performed and they had some problems. Upon arrival
I learned that the office was originally using Windows 98 and wanted better
security and file safety so they chose to upgrade to Windows NT. After the
upgrade, several computers had hardware problems, especially with mice
and keyboards. As it turns out, the hardware in question was interfacing
with the computer through the USB port. Unfortunately, Windows NT does
not support USB interfaces. Had this company researched carefully into the
upgrade, it would have realized this problem before deciding on performing
the upgrade.

Planning the Upgrade

Prior to upgrading a server, you should perform the following tasks:
 Review any technical documentation.
 Choose components.
 Schedule the upgrade.
 Perform a full backup of the server.
 Document the server’s current settings.
 Create a backout (rollback) plan.
 Test the upgrade.
 Schedule for server downtime.

Copyright ©2002 SYBEX, Inc., Alameda, CA

284 Chapter 9  Upgrading and Maintenance

Reviewing Technical Documentation

Most upgrades will come with some form of technical documentation
outlining the upgrade process. Reviewing this information will help you
familiarize yourself with the upgrade process before it is actually performed
and alert you to any issues that you need to be aware of. This includes all
manuals, last minute notifications, and all Readme files that were supplied
with the upgrade components. Within this documentation will be informa-
tion regarding installation processes, possible issues, and customer support.
Depending on the type of upgrade being performed (software, hardware,
or operating system), specific steps will have to be followed to ensure proper
operation and the desired outcome.
Most websites will provide information on upgrade procedures. Take
the time to browse through the information as well as check to see if there
are any newsgroups available or known issues or forums pertaining to the
upgrade. Most hardware manufacturers and operating system manufactur-
ers will have some area on their website where users can share information
or post questions. Chances are you are not the first network administrator to
perform the upgrade and others may have posted information or advice.

Choosing Components
In deciding on an upgrade, be sure to take the time to research numerous
vendors and manufacturers. The multitude of products and manufacturers
may seem overwhelming, but it will give a clear idea of the options and price
range available. During this time, comparison to the components currently
installed in the server will prevent any incompatibility. Most hardware and
software manufacturers have extensive documentation on their websites and
will share potential problems as well as incompatibility and known issues.
Once components have arrived, the next step is to confirm all parts are
accounted for and are not damaged. Copies of invoices and order forms
should be verified and then filed into a documentation folder. Should a time
come when confirmation or verification of parts is needed, referral to these
invoices will be of great assistance.

Schedule the Upgrade

Once the upgrade compatibility is confirmed and all known issues are
dealt with, a timeline on the upgrade can be estimated. This will include the
process of ordering the proper equipment, shipping, installing, testing, and
configuring the upgrade. This process will also include any “red tape” that

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment 285

may have to be taken care of. For example approval of the upgrade from
management or a board of directors may have to take place before the
go ahead.
Scheduling the upgrade entails determining what upgrade steps are going
to be completed at what times and who is responsible for performing them.
The plan that is developed will of course depend on the type of upgrade that
is being performed. A simple upgrade can be scheduled to occur during a
span of a few hours while a more elaborate upgrade may be scheduled to
occur over a span of several days (even months in some cases). Either way,
a schedule of an upgrade organizes the procedures so technical staff—and
maybe nontechnical staff—know what is going to occur and when it will
occur. Depending on the business policies and how relaxed they are, your
schedule may just include the date and time that the upgrade will occur or it
may include details such as when the hardware and software will be pur-
chased, when the testing phase will begin, and when users will be notified of
server downtime.

Talk to any experienced network administrator and one of the first things
they will recommend before making any changes to a server is to perform a
full backup. This means backing up the entire server. Your backup should
include a backup of the operating system, any applications, and all data
stored on the server.
Once you’ve performed a full backup, you may think that you are ready
to upgrade, but there have been many instances where a network adminis-
trator has backed up data only to find that it cannot be restored. So once
again, to be on the safe side, do a trial run and test the restore. You will want
to have a test machine on which to run the restore, just to make sure the
whole thing works. If you do not have a test machine available, it’s time to
get one.

Document the Configuration Settings

Chances are that your server has unique configuration settings, whether it is
OS settings or application and service settings. Documenting this informa-
tion can save you hours of work. If by chance the upgrade fails and you are
unable to restore your server from a backup tape, your only option will be
to perform a clean install. Having this information documented will make
your job easier when it comes time to reconfigure.

Copyright ©2002 SYBEX, Inc., Alameda, CA

286 Chapter 9  Upgrading and Maintenance

Create a Backout Plan

You may often hear a backout plan referred to as a rollback plan. A backout
plan is exactly that. Knowing when the upgrade has failed and how to get the
system back to its previous state. The backout plan should define at what
point in the upgrade process you need to stop and begin attempting to
recover the server. It should also outline the steps that will need to be com-
pleted to bring the server back to its original state. Again depending on the
type of upgrade, the plan may be as simple as replacing the new hardware
with the old.

Test the Upgrade

In an ideal world you should be able to test the upgrade plan in a controlled
test environment before performing the upgrade in the production environ-
ment. In that ideal world, you could test the upgrade several times. The test
environment should mirror the actual production environment. If at all pos-
sible, the test server should be configured exactly the same way as the server
to be upgraded, with identical hardware and the same configuration settings,
applications, and services. Doing a test run of the upgrade will determine
first of all whether or not it will succeed and also alert you to any minor (or
major) problems that might arise and how they can be remedied. Again,
unless the server is an exact replica of the server to be upgraded, the results
of your test run might not be quite accurate. However, it will give you some
hands-on experience with the upgrade process.

Always Read the Instructions

It is interesting how, after performing many expansion card upgrades,

we fall into the trap of thinking that they are all the same. The usual process
involved removing the cover, installing the card, replacing the cover,
turning on the computer, and finally letting the Plug and Play wizard locate
and install the card and the software. The reality is that this is a common
routine, but not the rule. I have encountered expansion cards that required
the software to be installed first, then the actual hardware installed second.
If you installed the hardware first and the software last, the card would
not work.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Assessment 287

Scheduling Server Downtime

Server availability is very important. During the upgrade you want to make
sure that server downtime is minimal. Again, the type of upgrade that you
are performing will determine the length of time that a server will be unavail-
able. In an ideal situation, there should be another server to take the place of
the one being upgraded while it is offline, but in instances where you have
a small office with a single server, this may not be feasible. However, if the
server is running services that are critical to the operation of the network, a
stand-in server may be necessary.
Both ways, server downtime still needs to be scheduled and users notified.
Users tend to understand that upgrades need to be performed as long as it
does not cause too much of a disruption to their workflow. When planning
to take a server offline, try to schedule it during non-peak hours; generally
this means performing an upgrade in the evening or weekend. This way the
impact on users will be lessened.
Scheduling server downtime also includes notifying users. This can be in
the form of an e-mail message telling users when the server will be down,
how long it is expected to be down, and any impact this might have on them.
You may also want to send a message to all users connected to resources on
the server just before the outage occurs so they have time to save their work
and disconnect from the server. Figure 9.3 shows an example of a console
message sent from a server running Windows 2000.

FIGURE 9.3 Windows 2000 console message

It’s a good idea to send out multiple messages to your clients warning
them that the server is coming down. Invariably though, after you bring the
server down, you will get a dozen phone calls from frantic employees that
either didn’t pay attention or didn’t save their files.

Copyright ©2002 SYBEX, Inc., Alameda, CA

288 Chapter 9  Upgrading and Maintenance

After the Upgrade

Obviously, if you are performing an upgrade you are making changes to the
configuration of a server. Preplanning is of the utmost importance, but what
about after the upgrade is complete? It is generally good practice and com-
mon courtesy to document any changes that have been made to the server
and keep the information in a log somewhere secure. If you are the only net-
work administrator, then obviously you are the one who monitors changes.
But if you are not, it is good practice to document the exact changes that
were made. Nothing can unnerve an administrator more than sitting down
at a server only to find significant changes and not know when they were
performed or what was done. Not only is providing documentation a com-
mon courtesy, but it also aids the next administrator in troubleshooting any
problems that may occur after the upgrade has been completed.

Risk Assessment
Any time you perform a server upgrade, there are risks involved that
you need to be aware of beforehand. So during the preplanning phase, one
of the tasks you need to complete (and this can be done when creating a
backout plan) is to perform a risk assessment. A risk can be defined as a
potential problem that may arise during the upgrade. A risk assessment
entails identifying the possible risks of the upgrade and having a contingency
plan in place should they occur. As an example, what would happen if you
upgrade Windows NT to Windows 2000 and the server’s NIC isn’t on the
Windows 2000 HCL? Your overall goal in performing a risk assessment is to
identify the risks, eliminate them if possible, or find ways to eliminate their
impact should they occur. Part of your risk assessment plan should be to
determine how critical the server is to the day-to-day operation of the busi-
ness and how much downtime can be afforded if a problem does arise.
Part of addressing the potential risk factor is determining if the upgrade is
necessary. Do the benefits of the upgrade outweigh the potential risks? It is
not only unnecessary, but also unwise, to install every available update that
you can find. At times, upgrades will conflict. It is very common to see
numerous versions of the same update released over time. Always check the
date on the upgrades to ensure that you are not downloading an outdated

Copyright ©2002 SYBEX, Inc., Alameda, CA

Risk Assessment 289

version. Other updates are designed specifically for an operating system,

network type, or version. If you are not using that specific network type
(for example), then the patch will not be useful.
In dealing with risks to operating system functionality, Microsoft intro-
duced a new feature with Windows 2000 called driver signing. Driver
signing identifies drivers that have been tested and certified to work properly
with Windows 2000. When drivers are installed that are not part of the Win-
dows 2000 driver signing, you will be prompted that they may not perform
as expected. This feature helps to eliminate the potential for unexpected
results with driver operations and potential risks to the operating system.
Microsoft also uses an HCL (hardware compatibility list). This list includes
vendors and products that were certified to operate within the Windows
operating system trouble free. Other devices may work, but there is a poten-
tial for resource or driver issues. Driver signing improved on the HCL by
alerting the installer to the potential risk.
Resource conflicts are another potential area of risk. In a mixed expan-
sion card environment, both PCI and ISA cards can be used. ISA card
resources are configured through jumpers while PCI card resources are
controlled through the BIOS and operating system. The potential for
resource conflicts becomes high in this environment. As discussed earlier,
Windows 2000 offers a view of shared resources and forced hardware
resources in the Computer Management utility. This will assist in locating
and dealing with resource problems.
With any upgrades, the potential for security holes and breaches also
becomes an area of risk. Careful planning during the installation and assess-
ment process will help prevent these potential risks. For example, the instal-
lation of a dedicated Internet connection exposes the server and network to
Internet-based threats such as hacking. Installing protective software such as
a firewall will reduce this risk.
Your risk assessment should include the following information:
 Identify the risk factors associated with the upgrade—the type of
upgrade you are performing will obviously determine the number of
risks involved and their potential impact (high or low impact).
 Determine the impact the risk factors may have. Once the potential
problems have been identified, assess each one to determine what may
happen if it occurs. What impact will it have on the users’ workflow
and the upgrade schedule? Are there any costs associated with the
risk and what effect will it have on the server?

Copyright ©2002 SYBEX, Inc., Alameda, CA

290 Chapter 9  Upgrading and Maintenance

 Determine the likelihood of each risk factor occurring—Ask yourself

or any other IT staff involved in the upgrade what the chances are of
the risk actually occurring.
 Develop a mitigation strategy—Once all the risks have been identified,
your last step in performing your assessment will be to determine what
steps can be taken to eliminate the likelihood of a potential problem
occurring or finding ways to lessen the impact (especially for those
problems that are likely to occur).

This is one instance where being a little on the paranoid side can be a positive
thing. Always plan for the worst, even though the chance of the worst actually
occurring is minimal. In the end you will feel much more confident when it
comes time to perform the upgrade.

Server Availability
Your ultimate goal is to make network servers available 100 percent
of the time (this is referred to as server uptime) or when users need to have
access to them. If a business is up and running 24 hours a day, 7 days a week,
your network servers will need to be available 100 percent of the time.
Although the goal is 100 percent uptime, in some cases this may be impos-
sible to achieve (take for instance a business with a single server). Obviously
the uptime required by businesses will vary depending on the applications,
services, and data being stored on the server and how critical they are to the
company’s day-to-day operation.

How Much Downtime?

This is a question you will need to ask yourself, other IT staff, and users on
the network before performing an upgrade (particularly if it is a major
upgrade). How much server downtime can the business actually afford? This

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Availability 291

will depend on the role the server plays on the network and how critical it is.
Take for example a file server that is accessed occasionally by users on the
network. In this case, taking the server offline will have little impact on the
users. If another server is available on the network, the files can easily be
duplicated and made accessible during the time the server is offline. What if
the server you are upgrading is a domain controller? Taking this server
offline for any amount of time can have a profound effect on the network
and will mean configuring another server to take its place in the meantime.
When determining how much downtime can be afforded, begin by assess-
ing what role the server plays on the network. Document all the services,
applications, and resources hosted on the server and how they apply to the
operation of the business. Anything that is critical to its operation may have
to be duplicated onto another server while the upgrade is done. Consider a
server that is leasing IP addresses to users on the network. If this server were
to be unavailable for a period of a week while an upgrade is performed, it
would have a major impact on what users could do on the network. The
service would then need to be duplicated onto another network server.

Increasing Availability
Increasing the availability of a server will depend on the environment in
which you are working. If you have access to multiple servers, then it is a
matter of duplicating those services or applications onto another one. The
following list includes some suggestions for minimizing the impact of server
downtime, thereby increasing server uptime.
 Schedule upgrades to occur during off-hours making the server
available when users need access to it.
 If possible, configure a server to take the place of the one being
upgraded while it is offline.
 Take advantage of clustering technology for critical servers.
 Take the time to complete the pre-upgrade tasks so you are prepared
for any potential problems that might cause the upgrade to take longer
than predicted.

Copyright ©2002 SYBEX, Inc., Alameda, CA

292 Chapter 9  Upgrading and Maintenance

This is one instance where clustering servers plays an important role. As you
recall, when servers are in a cluster configuration, one is waiting to take over
the workload of another if it goes offline. This includes any planned outages
for upgrade and maintenance procedures. The resources on one server can
failover to another server if the cluster and downtime is decreased to a matter
of seconds (users probably won’t even notice). You can then take the server
off the network and perform the necessary upgrade. Once the upgrade is
complete, the server can reassume its workload and all is well. Clustering is
probably not necessary for servers that play a minimal role on the network.
But for businesses that rely heavily on web servers for e-commerce, it can
essentially mean millions of dollars.

Performing the Upgrade

Once you have verified that all the parts for the upgrade have arrived, you
can begin the process of performing the upgrade. Based on the plan devel-
oped earlier, steps toward performing the upgrade can be carried out in the
carefully preplanned order. This will include a final reminder to all affected
clients that an upgrade will be performed. This is also the stage where a full
backup should be done. Performing a full backup will ensure that if a prob-
lem should occur, the data would be safe. Remember to also verify the
backup before continuing with the install.

Backups and verification are covered in depth in Chapter 14, “Backups.”

If the upgrade to be performed is hardware related then ESD (electrostatic

discharge) best practices must be adhered to. This includes the use of
antistatic bags, ESD wrist straps, and mats to prevent static charges from
damaging delicate hardware components. ESD damage is a serious threat to
computer hardware and can lead to hardware operation problems.

ESD best practices are covered in Chapter 15, “Disaster Recovery.”

Depending on the hardware upgrade being performed, there may be

an impact on the system’s power consumption. For example, installing
redundant components within a server could require upgrading of the power

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Availability 293

supply and/or the UPS. All servers should be connected to a UPS (uninter-
ruptible power supply). A UPS is nothing more then a battery system that
will maintain power to the server in the event of an electrical failure. The bat-
tery is connected between the server and the electrical outlet. When there is
a power failure, the battery system will run the server and any other devices
connected to the UPS for a period of time. Depending on the size and strength
of the battery system, a UPS may support minutes to hours of power.
Software is also available that will interface between the UPS and server
allowing for alerts and remote notification that the server is running off bat-
tery power. Another key feature that UPS software provides is a shutdown
option for the server. If the electrical power is not re-established before the
battery runs out, the software will alert users, close all software programs
that are running, and then power down the server operating system. This
prevents the hard shutdown that occurs when power fails and all open
programs including the operating system are abruptly shut down. Hard shut-
downs can cause loss of data as well as program corruption. If the UPS that
the server is currently using is underpowered, then an upgrade to the UPS
will also have to be made. UPS software also monitors the power load being
placed on the UPS battery. If the load exceeds the recommended battery
load, then the UPS should be upgraded. Anytime a hardware upgrade is
performed, the UPS load should be checked to confirm that it is capable
of handling the stress of the added component.
Working in conjunction with a UPS should be a surge protector. Some
manufacturers combine both a UPS and surge protector in one, but it is
advisable to purchase a dedicated surge protector. A surge protector has a
built-in circuit breaker that will trip should a spike of electricity arrive. This
prevents the momentary increase in electricity from damaging the computer.
Many surge protectors offer modem and network protection along with elec-
trical protection. This insures that any potential damage traveling down a
telephone or network line will be caught before it can reach the computer.
All sensitive equipment should be protected by a surge protector (including
computers, printers, fax machines, and networking products).
Software upgrades include patches, new software, upgrades to existing
software, and firmware. Patches are small programs that are installed to
repair or add features to existing software. An example would be a patch for
a database program to repair a known problem. Normally patches are sim-
ple to install, using an executable file that does all the work, but at times
patches will require you to manually install files or overwrite existing files.
New software includes any program, driver, or file being installed to the

Copyright ©2002 SYBEX, Inc., Alameda, CA

294 Chapter 9  Upgrading and Maintenance

server that has not previously been installed. Research to ensure compatibil-
ity with other software and the operating system is a must. Firmware (as
discussed in Chapter 3, “Motherboards and Processors”) is software that
controls hardware. Firmware updates on a server include SCSI controllers,
RAID controllers, tape drives, and CMOS BIOS. Although other hardware
devices have firmware, these four are the most common firmware updates.
When installing firmware updates, be sure to obtain the software from the
product manufacturer. The potential risks of component damage or inoper-
ability should be a deterrent enough to not download firmware updates from
just any Internet site. Manufacturer’s sites design and verify firmware updates
specifically for their products. By downloading from the manufacturer’s site,
you can be reasonably certain about the reliability of the firmware.
Operating system upgrades and updates normally require purchasing the
update on a CD-ROM or downloading it from the operating system website.
Similar to software updates, operating system updates include repairs to
known problems as well as enhancing features. Again, before installing any
update, confirm that it will be compatible with your existing hardware and
software. Windows 2000, for example, offers updates called service packs.
Two service pack updates have been released to deal with known operating
system problems. Operating system updates are available for Windows,
NetWare, and Unix based systems. Frequently, visiting the operating system’s
website will help keep you up to date on the current releases as well as their
role. New operating systems such as Windows 2000 also offer notification of
new patches and service packs. This alert can come in the form of an e-mail
or a popup notification in the system tray.

After the Upgrade

Once the upgrade has been performed, as outlined in your predetermined
plan, the server should be restarted (if needed) and observed. Observation
will include visually checking for problems. This includes the POST (power-
on self test), any beep or visual error codes that appear, or error messages
generated. Any abnormalities noted at this point will need to be immediately
addressed. Even if the error message generated does not appear to have any
immediate effect on the server operation, it does point to a misconfiguration.
With time this misconfiguration might cause a problem.
Windows 2000, for example, has a computer management feature, which
is accessed by clicking the right mouse button on the My Computer icon on
the desktop and selecting Manage. Figure 9.4 is a screen shot of the com-
puter management utility.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Availability 295

FIGURE 9.4 Windows 2000 Computer Management

With this management feature you can verify software and hardware
operations. The event viewer will assist in identifying potential problems with
applications, Internet Explorer, security, and system. Figure 9.5 shows errors
generated by the system log. Notice how errors are shown as an X in a circle
graphic in the right pane, while warnings are an exclamation point in a triangle.

FIGURE 9.5 Windows 2000 Event Viewer

Copyright ©2002 SYBEX, Inc., Alameda, CA

296 Chapter 9  Upgrading and Maintenance

Computer Management also will analyze and display system informa-

tion. This includes a system summary, hardware resources, components,
software environment, Internet Explorer, and applications. System infor-
mation provides a clear picture of how the computer is currently running.
Any potential issues can be identified. Figure 9.6 shows the hardware
resources tab expanded. Here information on system resources (IRQ,
DMA, I/O, and system memory) can be seen as well as conflicts/sharing
and forced hardware. Forced hardware is hardware that has resources set
by the user rather than letting the operating system or BIOS determine the

FIGURE 9.6 Windows 2000 Hardware Resources

A final area in Computer Management, which will assist in upgrade

monitoring, is the Disk Management tab. Disk Management will provide
you with information on hard disks, optical drives, and removable storage
(such as a Zip drive). Information on drive status, size, free space, percentage
used, drive letter, file system, layout, volume, and type can be identified here.
Figure 9.7 shows the Disk Management tab.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Server Availability 297

FIGURE 9.7 Windows 2000 Disk Management

A final step to performing an upgrade includes removing any old drivers

or software to prevent potential conflicts. Although obvious, the replaced
hardware should also be removed. At times this simple step is overlooked.
Upgrading a 28.8 modem with a faster 56K modem should include removing
the old 28.8 modem. Not only will removing the old modem prevent poten-
tial conflicts with the new modem, but also free up system resources. Remov-
ing old software and drivers will also free up hard drive space.
Once the upgrade has been performed and verified as successful, access to
the server can be restored. This process should be done selectively to ensure
that a smooth transition occurs, as well as verifying that the upgrade has also
been successful from the client side. Ideally, access to the server should be
granted to one group at a time and verified. For example, if there is a sales
group, management group, supplies group, and administration group, access
should be given to one group at a time and confirmed that the upgrade has
been successful before granting access to the second group. Through this
process any potential problems can be isolated and dealt with accordingly. If
access to the entire network is given all at once, any group-related problems
will be more difficult to isolate—and you can expect a bombardment of calls
from all the affected groups.

Copyright ©2002 SYBEX, Inc., Alameda, CA

298 Chapter 9  Upgrading and Maintenance

Once everything is returned to normal and the server is fully functional
again, a detailed documentation of the upgrade should be created. This
information can assist at a later date should the need to retrace steps occur.
Also, any issues that came up as a result of the upgrade should be outlined.
This documentation should be kept in a safe location and include the plan-
ning stage as well as all shipping and order invoices. Be sure to outline the
upgrade procedures clearly, including screen shots of error messages and
detailed steps to rectify the errors.

Unsuccessful Upgrades
F or those of you who have experience performing server upgrades,
you know that there are times when the upgrade will fail, regardless of how
much time and preparation went into planning. Hardware and software
upgrades can be unpredictable and when it comes to performing upgrades,
sometimes failure is an option. So at some point you may have to stop trying
to fix a failing upgrade and begin the recovery process.
The effects of a failed upgrade can certainly be minimized by your pre-
planning tasks. This is when you can refer to the backout plan that you
developed. The backout plan is going to tell you how to recover from the
failed upgrade and restore the server to its original state (this maybe through
restoring from a backup tape or reverting back to the original hardware).
The backout plan should also give you an indication as to when it is time to
stop the upgrade and begin restoring the server. For example, if you have
scheduled the server to be unavailable for a specific amount of time and are
reaching the time when it should be brought back online, it is probably time
to stop trying to fix the upgrade and start trying to bring the server online
again. As difficult as it may be, especially after the time and effort that is put
into planning for the upgrade, at some point you will need to make the deci-
sion to cut your losses and start your restore.
One of the most important steps you can take after a failed upgrade is to
document the entire process. All is not lost from a failed upgrade and it
should be looked upon as a learning experience. Document as much of the
upgrade process as possible, such as when it failed, the steps you took to fix
the problem, error messages that were generated, and any other information
that will help you to determine why the upgrade failed. Chances are you will

Copyright ©2002 SYBEX, Inc., Alameda, CA

Maintenance 299

attempt to perform the upgrade again in the future and this information
will useful for troubleshooting.

Planning Future Expansion and Availability

The hardest thing to do is predict the future. When designing and building a
server, you have to try to look to the future. What will the requirements be?
What software will be required? What stress will the software place on the
hardware? A server should be built with the room for hardware expansion,
including expansion slots, room for hard drives, and redundant components.
Even if current financial situations prohibit purchasing of redundant com-
ponents, purchase a server case that will support the addition of redundant
components. Later, when finances are available, redundant components can
be purchased and added.
A key component to build on is the motherboard. Server motherboards
(as covered in Chapter 3) should support multiple processors, large amounts
of RAM, and expansion cards. Having a motherboard that will allow you to
expand will pay off in the long run with ease of expansion and upgrading.
Server upgrading oftentimes is a consistent process. From the beginning
point of assessment, through the component research, compatibility veri-
fication, ordering and installing, and finally returning access back to the
users, it often takes a great deal of time. By the time the process is complete,
there is another area that will require upgrading attention (software, hard-
ware, or operating system).

R egular maintenance on a server ties in closely with upgrading.
Regular maintenance includes monitoring server functionality as well as
performing regular routines to maintain server operations.

Proactive Maintenance
Proactive maintenance is performed to prevent problems from occurring. An
example of a common proactive maintenance activity is cleaning backup tape
drives regularly. By doing this simple activity you can prevent backup failures
due to dirty or magnetized heads on the tape unit.

Copyright ©2002 SYBEX, Inc., Alameda, CA

300 Chapter 9  Upgrading and Maintenance

Other forms of proactive maintenance include monitoring environmental

conditions such as climate (heat and cold) and dust. No computer will per-
form well in environments that are too hot. By monitoring and maintaining
a favorable temperature for your server to work in, you can prevent issues
associated with overheating and premature hardware failure. Dust will accu-
mulate over time in any environment. As these particles collect, they form a
blanket on computer components, which in turn will increase the tempera-
ture within the computer. Dust particles will also, over time, collect on cool-
ing fans and can cause premature fan failure. This has often been the reason
for power supply failure. As the fan in the power supply turns, it draws air
into the power supply and computer. Dust is also drawn in, which accumu-
lates on the fan motor spindle. This will cause the fan motor to squeal and
eventually fail. How to prevent this issue is to use compressed air regularly
to clear dust from the computer. Compressed air comes in cans with a straw-
like attachment. This attachment allows you to position and direct the
high-pressure air into areas that are not easily accessible.

You should only use compressed air designed for computer use. Air compres-
sors will use pressurized air that may contain contaminates such as oils.
Using suction devices such as vacuum cleaners can cause harm by actually
ripping delicate parts off.

Air filters can also be installed over the fan intake grill. These filters trap
the dust particles before they can enter the fan assembly. Since servers nor-
mally contain several fans that move air throughout the server, the use of fil-
ters can be a good idea. These inexpensive devices can prevent dust buildup,
especially in areas that are not readily accessible for cleaning. However, if
you’re going to use them, make sure to change them frequently. The last
thing you want is for them to get clogged up, causing the system to overheat.
Probably the most common known area of proactive maintenance is in
virus protection. Viruses are a threat to any computer, and with new viruses
emerging weekly, they should be taken seriously. Virus protection in a net-
work can be based individually on each computer, or server-centric. Server-
centric virus program are called a virus protection suite. This software offers
centralized protection and updates. The real benefit of a virus protection
suite occurs when the time to update the virus definition comes. A virus
definition is a list of the known viruses that is used by the virus protection
engine to monitor your computer. If the virus program is individually based,

Copyright ©2002 SYBEX, Inc., Alameda, CA

Maintenance 301

this definition will have to be updated on each computer in the network

individually. Virus protection suites are updated on the server, which then
monitors the rest of the network. Depending on the size of your network,
having to update just the server as opposed to every system on the network
can be a huge difference in time. Regardless of the type of virus protection
that you use, it is essential that there is some form of protection in place
and that it is updated regularly. A virus on a server can lead to severe con-
sequences, including data loss and loss of productivity. Imagine a company
such as eBay losing a few days of business because a virus brought its servers
down. It would be a financial nightmare.

Baselines, Monitoring, and Thresholds

Another element of proactive maintenance involves carefully monitoring
your server’s performance. At the beginning of the chapter we discussed the
use of monitoring tools for server performance as it relates to the need for
upgrading hardware and software. Server monitoring is also an effective
means of maintaining a server. Monitoring can provide advanced warning of
performance issues before they lead to larger problems.

Baselines are measurements of server performance. This measurement is
taken over a period of time to determine how well a server will handle appli-
cation and stress loads. Based on the information gathered, a baseline of
performance can be set. This will help distinguish acceptable server perfor-
mance from unacceptable performance when the server is operating under
normal and heavy loads.
One of the key tools used in creating baselines in Windows 2000 is the
System Monitor. As discussed earlier in this chapter, the System Monitor
provides a means of visually monitoring how a server is doing. Using the Sys-
tem Monitor to record server performance a baseline is easily created. This
Computer Management feature will even alert you if the performance dips
below the established baseline.

Thresholds are set values similar to baselines; however, with thresholds,
there is an acceptable range of values. This set minimum and maximum
range conforms to a safe operating limit for the computer component.

Copyright ©2002 SYBEX, Inc., Alameda, CA

302 Chapter 9  Upgrading and Maintenance

Thresholds are used to monitor temperature and electrical signals in a UPS.

Should the electricity coming in from the outlet spike or dip below the
threshold ranges, then the UPS will react accordingly. Thresholds can also
be set to monitor resources within the server. CPU load, RAM usage, and
free hard drive space are a few of the commonly monitored areas.

T his chapter began with defining assessment. Understanding the need
to assess, both as it relates to the need for upgrading as well as system mon-
itoring, is the key to this chapter.
Assessment in the Windows 2000 environment is done through the use
of Computer Management program. Computer Management provides a
means of assessing hardware and software and their interactions with the
operating system.
Upgrades can take on the form of software, hardware, or operating
systems. Although each upgrade is unique, the stages involved in performing
the upgrade can be broken down into three key areas: planning the upgrade,
performing the upgrade, and managing the system after the upgrade. Plan-
ning the upgrade includes determining due dates; researching products and
vendors; confirming compatibility with existing hardware, software, and
operating systems; verifying component delivery; and thoroughly reading all
documentation. Performing the upgrade includes performing a full backup,
being aware of ESD, and actually doing the upgrade as planned. Managing
the system after the upgrade covers making sure that the upgrade was per-
formed successfully, verifying that there are no negative effects as a result
of the upgrade, removing any old hardware or software, returning access to
clients, and finally documenting the details of the upgrade.
Server availability and upgrade failures also need to be considered before
performing an upgrade. When planning for an upgrade, you will need to
assess the business requirements as well as the role the server plays on the
network to determine how much downtime can be afforded. Upgrade fail-
ures are bound to happen, and this is where your preplanning will be useful.
Creating a backout plan prior to the upgrade will help you to recover the
server if indeed the upgrade is unsuccessful.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Exam Essentials 303

Windows 2000 Computer Management also offers a means of visually

monitoring forced hardware and resource conflicts, which are potential risks
to stable server operation.
Planning for future expansions and availability focuses on building a
server to grow with. If planned correctly a server should be upgradable to
meet current as well as future demands. This includes support for redundant
components, multiple processors, and more RAM.
Maintenance is an important element of stable server operation.
Proactive maintenance involves performing tasks to reduce the possibility
of a problem. This includes tasks such as regular tape drive cleaning, using
compressed air to clear dust off internal components, and filters to prevent
dust from accumulating on fan motors.
The use of a surge protector and UPS will protect against potential elec-
trical issues.
Virus protection suites also fall under proactive maintenance. Using a
regularly updated virus program will ensure that the threat posed from
viruses can also be minimized.
Baselines are set values by which performance can be measured. Using the
performance monitor in Windows 2000’s Computer Management program
will assist in creating and monitoring server performance. With an estab-
lished baseline of performance, potential bottlenecks and problem areas can
be identified before they create a serious problem.
Thresholds are similar to baselines, but they monitor more than one
variable. Thresholds identify a range of acceptable performance. Should this
range be exceeded (either higher or lower), an alert is sent out.

Exam Essentials
Know the benefit of monitoring tools. Monitoring provides a means of
watching and assessing server performance.
Know how to use Windows 2000 System Monitor. Be familiar with
the available counters in System Monitor as well as how to add counters
and how to monitor performance.
Know the difference between counter logs, trace logs, and alerts.
Counter logs record information gathered by the performance monitor,
trace logs track changes and actions, and alerts send out a notification
when a threshold is reached.

Copyright ©2002 SYBEX, Inc., Alameda, CA

304 Chapter 9  Upgrading and Maintenance

Know what a bottleneck is and be able to identify where a bottleneck is

located. Bottlenecks are areas where performance is hindered by poor
performance. Windows 2000 offers Computer Management software to
assist in locating bottlenecks.
Perform a risk assessment. A risk assessment identifies all the potential
problems that may occur during the upgrade, the likelihood of them
occurring, and the steps you will take to alleviate their impact.
Be able to determine the requirements for server availability. Server
availability is the amount of time the server is available when users need
access to it. Business requirements and the role the server plays will deter-
mine the required uptime for a server.
Know the steps to planning an upgrade. These include developing an
upgrade plan, researching vendors, verifying compatibility, confirming
parts on arrival, and reading all documentation.
Know the steps in performing an upgrade. Be aware of ESD issues,
perform a full backup, and alert users to the upgrade time.
Recover from a failed upgrade. At some point you will have to stop
trying to fix a failed upgrade and begin the recovery process. Use the
backout plan that you created during the preplanning phase to restore
the server.
Know what a UPS is and how it works. Assess the stress load on a UPS
after an install and determine if it will require an upgrade also.
Know the use of a surge protector. Be able to identify the benefit of
using a surge protector to protect against electrical, telephone, and
network threats.
Know what a software patch and firmware update are. Software
patches help fix issues with software problems; and a firmware update is
an update to software that is specifically embedded in the hardware.
Know the steps to confirming a problem in an upgrade. This includes
identification of the POST, error codes, and error messages.
Know how to create documentation after the install is complete.
Include all steps performed, any problems encountered as well as steps
and solutions taken to resolve them.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Key Terms 305

Know how to assess potential risks both to computer and business.

Risks include any threat that may cause a problem including resource
management, upgrades, Internet connections, and virus.
Know steps in proactive maintenance. This includes regular cleaning,
dust removal, controlling environmental variables, and using a virus
protection suite.
Be able to establish baselines and thresholds. Baselines are measure-
ments of server performance; thresholds are a set range of values. Both
can be set within the computer management feature of Windows 2000,
or through the use of third-part software.

Key Terms
B efore you take the exam, be certain you are familiar with the
following terms:

alert monitoring
assessment monitoring agents
baseline patches
bottleneck POST (power-on self test)
Counter Logs Monitor server performance
driver signing surge protector
ESD threshold
firmware Trace Logs Monitor
hard shutdown upgrading
HCL (hardware compatibility list) UPS (uninterruptible power supply)
log files virus protection suite

Copyright ©2002 SYBEX, Inc., Alameda, CA

306 Chapter 9  Upgrading and Maintenance

Review Questions
1. Which of the following is not a means of assessing and maintaining
a server?
A. Establishing baselines

B. Reviewing monitoring agents

C. Reviewing log files

D. Visual cues

2. How often should assessment be carried out?

A. Once

B. Weekly
C. Monthly

D. Constantly

3. What Windows 2000 program allows you to assess computer

A. Diagnostic Monitor

B. Device Monitor

C. System Manager

D. System Monitor

4. What is an alert?

A. A notification of software shutdown

B. A means of setting a threshold on a counter

C. A visual reminder of a problem

D. An error code used by servers

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 307

5. What is a bottleneck?

A. A blockage within a software program

B. Locations where performance is hindered due to poor performance

C. The funneling effect of data into one program resulting in software

freezing up
D. A type of virus

6. Which of the following is not one of the three main types of upgrades
that can be performed?
A. Hardware

B. Software

C. Operating system
D. Virus

7. What does the acronym ESD stand for?

A. Electrostatic discharge

B. Electronic system damage

C. Electrical system damage

D. Electron static discharge

8. What does the acronym UPS stand for?

A. Unserviceable power system

B. Uninterruptible power supply

C. Unusable power supply

D. Unstable power system

Copyright ©2002 SYBEX, Inc., Alameda, CA

308 Chapter 9  Upgrading and Maintenance

9. What type of backup should be performed prior to an upgrade?

A. Differential

B. Partial

C. Incremental
D. Full

10. What is a hard shutdown?

A. An abrupt, unplanned loss of power

B. Manually shutting down a computer

C. Completely shutting down a computer

D. Forcefully shutting down software that is not responding

11. What are patches?

A. Small programs that repair or add features to programs

B. Programs that are installed on a server

C. A type of virus

D. Updates to hardware components

12. What is a firmware update?

A. A type of patch

B. Software updates for hardware components

C. A branch of software used on a server

D. A dedicated software program for hard drives

13. Operating system updates are available for which of the following
server operating systems?
A. Windows
B. NetWare

C. Unix

D. All of the above

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 309

14. What is the purpose of establishing a server baseline?

A. A baseline will help to determine what is a normal level of perfor-

mance for a server.
B. A baseline will assist you in determining if the planned upgrade
will actually work.
C. A baseline will determine how long the server will be down during
an upgrade.
D. A baseline will determine the amount of server availability that a
business requires.

15. Mike is doing a software upgrade to a server. He’s done installing the
software and testing the server and is just about ready to put the unit
into production. What one last item should he take care of before
making the server a production server?
A. Download and apply any service patches.

B. Download and apply any NOS upgrades.

C. Obtain a baseline of the computer.

D. Add any necessary service tools that weren’t initially installed.

16. What is one way you can ensure that you will be able to recover a
server from a failed upgrade attempt?
A. Read the technical information associated with the upgrade.

B. Back up the server before beginning the upgrade.

C. Test the upgrade in a controlled environment.

D. Identify all the associated risks.

17. Users on the network have been complaining that one of the servers
has been very slow lately and seems to be getting worse as the weeks
progress. What is your first step in alleviating the problem?

Copyright ©2002 SYBEX, Inc., Alameda, CA

310 Chapter 9  Upgrading and Maintenance

A. Perform an upgrade at once.

B. Leave the server as it is and see if the problem persists.

C. Take the server offline immediately.

D. Start monitoring the server’s performance and comparing it

against your baseline.

18. How do you access Computer Management in Windows 2000?

A. Through control panel

B. By right clicking on My Computer

C. Through the My Documents folder

D. In the Program Files folder

19. What does POST stand for?

A. Pre-operating system test

B. Power-off system test

C. Power-on self test

D. Pre-on self test

20. What is the final step in upgrading?

A. Turning the server back on

B. Granting user access to the server

C. Verifying that the install was successful

D. Documenting the install

Copyright ©2002 SYBEX, Inc., Alameda, CA

Answers to Review Questions 311

Answers to Review Questions

1. D. Visual cues are not a commonly used method of assessing and
maintaining a server as normally there is no user constantly sitting at
a server.
2. D. Assessment should be an ongoing process by which server perfor-
mance is monitored. This will ensure that any gradual performance
degeneration can be dealt with.
3. D. System Monitor is the Windows 2000 based program that allows
you to assess computer performance.
4. B. Alerts are a means of establishing a threshold on a counter. When
the threshold is reached an alert is sent out to notify of the problem.
5. B. Bottlenecks are spots where performance is hindered due to
poor performance. This can commonly include RAM, processor,
or software.
6. D. Hardware, software, and operating system are the three major
types of upgrades that are performed.
7. A. ESD is the acronym for electrostatic discharge. In this context it
is the transfer of static electricity from human contact to computer
8. B. A UPS is an uninterruptible power supply. It provides battery
backup in case of electrical failure.
9. D. A full backup should be performed before undertaking any
upgrade. This will ensure that, if there is a problem, all the data is still
safe on a backup.
10. A. A hard shut down is an unexpected, unplanned loss of power. It
can result in both data and software errors and should be avoided.
11. A. Patches repair or add features to existing programs. They are
located at software manufacturers’ websites.
12. B. Firmware updates are updates for hardware that contains a BIOS.
This includes RAID, SCSI, and motherboards.
13. D. Updates are available for all operating systems. They can be
located at the manufacturer’s website.

Copyright ©2002 SYBEX, Inc., Alameda, CA

312 Chapter 9  Upgrading and Maintenance

14. A. A server baseline is used to track server performance over time

and is later used for comparison to pinpoint bottlenecks.
15. C. Mike should run a baseline on the computer so he has a feel for its
pristine state characteristics. He will run performance-monitoring
software and capture the performance statistics off to a file for later
retrieval and review. This will aid him greatly should there be a prob-
lem with performance when the server goes into production.
16. B. Backing up the server’s operating system, applications, and data
will ensure that the server can be restored in the event that the upgrade
does fail. Remember to test the restore process before doing the
17. D. There is a fine line between a computer being slow and a com-
puter being broken. Compare it to the baseline and see if the users are
correct or are just thinking it’s slower. Once you have data, then you
can act.
18. B. You access Computer Management by right-clicking My Com-
puter and selecting Manage.
19. C. POST is the power-on self test. It is performed at system startup.

20. D. The final step in any upgrade should be documenting all the
details of the installation.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Chapter Hardware Updates



 1.2 Install hardware using ESD best practices (boards, drives,

processors, memory, internal cables, etc.)
 Mount the rack installation
 Cut and crimp network cabling
 Install UPS
 Verify SCSI ID configuration and termination
 Install external devices (e.g., keyboards, monitors,
subsystems, modem, rack, etc.)
 Verify power-on via power-on sequence

 2.6 Update manufacturer specific drivers

 3.3 Add hard drives

 Verify that drives are appropriate type
 Confirm termination and cabling
 For ATA/IDE drives, confirm cabling, master/slave and
potential cross-brand compatibility
 Upgrade mass storage
 Add drives to array
 Replace existing drives
 Integrate into storage solution and make it available to the
operating system
 Perform upgrade checklist, including: locate and obtain
latest test drivers, OS updates, software, etc.; Review FAQs,
instructions, facts and issues; test and pilot; schedule
downtime; implement using ESD best practices; confirm
that the upgrade has been recognized; review and
baseline; document the upgrade.

Copyright ©2002 SYBEX, Inc., Alameda, CA

 3.5 Upgrade BIOS/firmware
 Perform upgrade checklist including: locate and
obtain latest test drivers, OS updates, software,
etc.; review FAQs, instructions, facts and issues;
test and pilot; schedule downtime; implement
using ESD best practices; confirm that the upgrade
has been recognized; review and baseline; document
the upgrade.

 3.6 Upgrade adapters (e.g., NICs, SCSI cards, RAID, etc.)

 Perform upgrade checklist including: locate and
obtain latest test drivers, OS updates, software,
etc.; review FAQs, instructions, facts and issues;
test and pilot; schedule downtime; implement
using ESD best practices; confirm that the upgrade
has been recognized; review and baseline; document
the upgrade.

 3.7 Upgrade peripheral devices, internal and external

 Verify appropriate system resources (e.g., expansion
slots, IRQ, DMA, etc.)
 Perform upgrade checklist including: locate and
obtain latest test drivers, OS updates, software,
etc.; review FAQs, instructions, facts and issues;
test and pilot; schedule downtime; implement
using ESD best practices; confirm that the upgrade
has been recognized; review and baseline; document
the upgrade.

 3.9 Upgrade service tools (e.g., diagnostic tools, EISA

configuration, diagnostic partition, SSU, etc.)
 Perform upgrade checklist including: locate and
obtain latest test drivers, OS updates, software,
etc.; review FAQs, instructions, facts and issues;
test and pilot; schedule downtime; implement
using ESD best practices; confirm that the upgrade
has been recognized; review and baseline; document
the upgrade.

Copyright ©2002 SYBEX, Inc., Alameda, CA

 3.10 Upgrade UPS
 Perform upgrade checklist including: locate and
obtain latest test drivers, OS updates, software,
etc.; review FAQs, instructions, facts and issues;
test and pilot; schedule downtime; implement
using ESD best practices; confirm that the upgrade
has been recognized; review and baseline; document
the upgrade.

 4.5 Perform hardware verification

Copyright ©2002 SYBEX, Inc., Alameda, CA

A s network environments change, so must servers. Situations
will change and servers will need to be upgraded to meet the new demands
being placed on them. All too often servers are installed and configured and
then forgotten. The common phrase “if it is running fine then leave it alone”
does not apply to all situations, especially servers. As a server administrator,
you will undoubtedly need to perform a hardware upgrade to a server.
Chapter 8, “TCP/IP,” looked at the general steps to take when performing
any upgrade—hardware or software. Now we will look more closely at com-
mon hardware upgrades. Knowing how to upgrade hardware is important
for the Server + Exam and for your on-the-job success.

Before Upgrading
B efore you perform a hardware upgrade, you need to do a little
 Research the proposed upgrade to determine whether it is, in fact,
what is needed to improve the network performance.
 Read the documentation thoroughly before beginning the upgrade.
Make sure you are familiar with each step in the process.
 Make sure the hardware you choose is supported by the operating
system running on the server.
 Be prepared to comply with ESD (electrostatic discharge) best prac-
tices, including the use of antistatic bags, and ESD wrist straps and
mats to prevent static charges from damaging delicate hardware com-
ponents. ESD damage is a serious threat to computer hardware and
can lead to hardware operation problems.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Adding a Processor 317

 If the upgrade requires taking the server offline, plan a convenient time
to do this and notify your users in advance.

Adding a Processor
O ne of the most common components to be upgraded in a server is
the processor, whether you are upgrading a single processor or upgrading to
multiple processors.

In some cases you might be better off upgrading to a new server rather than
just upgrading the processor. This will depend largely on whether or not your
server’s motherboard supports current technologies.

The first thing to determine is whether a processor upgrade is going to

solve the performance problem—whether it’s the component that actually
needs to be upgraded. Using a tool such as Microsoft’s System monitor,
assess the processor’s performance under the current workload.
If you determine that a processor upgrade is necessary, you’ll need to do
some research on the motherboard itself. The type of processor you upgrade
to will depend on what your current motherboard can support. You can’t
just take any CPU and place it in the slot or socket on the motherboard
expecting it to work. Motherboards are limited to certain versions of CPUs.
So before you go out and spend the money on a faster processor, make sure
that the motherboard can support a faster processor. Examine the mother-
board documentation to determine the maximum speed the motherboard
The physical layout of the board (socket or slot) will also impact the CPU
upgrade. For example, you wouldn’t be able to use a socket 7 design with a
socket A design. Or if you plan to upgrade to Intel’s P4 chip, which comes
with a 423-pin design, the motherboard might have to be upgraded as well
(a 387-pin design will obviously not support a 423-pin chip). Supporting a
new processor might require a BIOS upgrade too (BIOS upgrades are dis-
cussed later in the chapter). Some motherboards will not work reliably with
the new processor until a BIOS upgrade is done.
Another aspect to consider is the cooling fan and heat sink currently used.
If you plan to upgrade the processor, make sure that the cooling fan and heat
sink will provide sufficient cooling. A processor upgrade will usually come

Copyright ©2002 SYBEX, Inc., Alameda, CA

318 Chapter 10  Hardware Updates

with a new fan and heat sink, but even if you need to buy them, the cost will
be small compared to possible losses from overheating. Fans and heat sinks
tend to have a short life span. Check with the manufacturer as well because
using a fan and heat sink that have not been tested with the processor may
mean that your warranty on the new CPU becomes void.
As with any other upgrade, begin by reviewing the documentation that
comes with the hardware for procedures on how to add the component and
perform a full backup of your server.

Performing the Upgrade

The following steps outline the general process in upgrading to a new
processor—the specific steps you take will follow the manufacturer’s
documentation. Make sure you adhere to ESD best practices.
1. After you’ve performed a full system backup and verified that it was
successful, turn off and unplug the server.
2. Remove the cover from the server and unplug and remove the CPU fan
and the heat sink (make sure you are properly grounded).
3. How you remove the old processor will depend on whether it has
a socket or slot design. With a socket design, release the lever to
remove the old processor. A slot design will probably use some form
of release mechanism on either side or both sides of the processor.
Either way, if you are unsure of how to remove the processor, consult
the documentation.
4. Install the new processor according to the documentation and attach
the CPU fan and heat sink.
5. After the new processor has been installed, a newer motherboard
should automatically detect its speed. An older motherboard might
not, so you will have to manually configure the BIOS or the jumper

Make sure that you correctly insert the processor. Be very careful to seat the
pins correctly before applying pressure; the pins bend easily. Make sure pin 1
on the new chip is plugged into the pin1 socket. Turning the server back on
while the chip is inserted incorrectly can permanently damage both the
motherboard and the processor.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Adding a Processor 319

The most common problem you will face when troubleshooting a newly
installed processor is the server’s failure to start. If this is the case, you will
need to put your troubleshooting skills to work and determine where the
problem lies. Start with these procedures:
 Check to see that the newly installed processor is properly seated in
the socket or slot.
 Check the documentation to verify that the jumper settings are prop-
erly configured.
 Remove the processor to make sure that none of the pins are damaged.
 Check whether a BIOS upgrade needs to be performed.
 Check the manufacturer’s support site to see if there are any known
If all else fails, you still have the old component to fall back on. Reinstall
the old processor—if the server starts, then the new processor is probably
faulty and needs to be returned to the manufacturer.

After you determine that the upgrade has been successful, start baselining
again to ensure that the new processor is meeting your performance expec-
tations as well as the required demands.

Multiple Processors
A processor upgrade might very well include upgrading from a single pro-
cessor to a multiprocessor or upgrading multiple processors in a system.
In today’s server environments, most servers are indeed multiprocessor sys-
tems. Your first consideration will be whether or not your operating system
and motherboard support multiprocessors.

The chip itself must also support symmetric multiprocessing (SMP).

Copyright ©2002 SYBEX, Inc., Alameda, CA

320 Chapter 10  Hardware Updates

Another consideration in multiprocessor upgrades is processor stepping.
Stepping is similar to version numbers: as updates are made to chips, the
version numbers change. You’ll want to consider processor stepping partic-
ularly when upgrading a single processor system to a multiprocessor one.
Mixing processor steppings does not always work well if it works at all. The
general rule of thumb is one stepping (revision) between CPUs is acceptable.
Information on stepping compatibility can be located from the manufac-
turer’s website. Try to purchase a chip with the same stepping, although if
you are dealing with an older chip this may be difficult. Some operating sys-
tems will be more tolerant when mixing steppings than others will be.

CPU Speed and L2 Cache

Two other considerations when upgrading a server to a multiprocessor sys-
tem are the CPU speed and L2 cache. The speed of the new chip must match
the speed of the existing chip. A CPU running at 200MHz cannot be paired
up with a CPU running at 700MHz. Also, the size of the L2 cache must
match. If the current CPU has a 250L2 cache, the processor to be added
must match this.

Adding Hard Disks

M ost servers can support multiple hard disks (also commonly called
hard drives), which are frequently upgraded in a server. One reason for
upgrading a hard disk is to gain increased disk space. A file server or a print
server can quickly run out of hard disk space, making an upgrade to a larger
hard disk or multiple hard disks necessary. An upgrade also could be done
to take advantage of faster data transfer speeds in newer hard disks. What-
ever the reason for the upgrade, there are certain steps that will need to be

Your first step in upgrading will be to determine whether you are dealing
with SCSI or IDE hard disks because this will impact which type of disk
you purchase and the procedure for installing it.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Adding Hard Disks 321

Upgrading SCSI Hard Disks

Several things need to be considered when upgrading or adding a SCSI
disk. Be aware that SCSI disks can give off plenty of heat; if you are installing
multiple SCSI disks, appropriate cooling needs to be considered. Make sure
that your system can provide adequate cooling for the number of disks you
install. Also, before going out and purchasing new SCSI disks, keep in mind
that they should match the speed of any existing SCSI disks in the system
(your slowest RPMs are also your fastest).
As with any SCSI device, a hard disk must be assigned a unique SCSI ID
(different from all other SCSI devices on the bus). The SCSI ID can be set
using the jumpers on the hard disk (refer to the manufacturer’s documenta-
tion). If there are already SCSI devices in the server, you will need to deter-
mine which IDs are available. If there is already a primary disk within the
server, set the SCSI ID to a number that is higher than that of the boot disk.
If you are replacing an existing SCSI disk, the new disk can usually be con-
figured with the same SCSI ID as the one being replaced.

Some SCSI IDs will not be set using the jumper pins. For example, if you are
adding a SCSI disk to an existing hardware RAID implementation, the RAID
system itself will assign the disk an ID.

Termination also needs to be considered. Most SCSI devices are now self-
terminating but you will need to consult the manufacturer’s documentation
to determine this (there may be a termination jumper pin that needs to be
set). Usually the SCSI adapter and the last disk on the chain are terminated.

Troubleshooting SCSI Disks

The following are some general troubleshooting tips for when problems arise
with the installation of a SCSI disk. Consult the manufacturer as well
because they will more than likely have an extensive knowledge base of
known issues.
 Verify the SCSI ID has been set correctly. If you encounter a problem
referred to as phantom disk, where the disk can be seen but not read
from or written to, then the new disk was probably configured with
the same SCSI ID as another device.

Copyright ©2002 SYBEX, Inc., Alameda, CA

322 Chapter 10  Hardware Updates

 Verify that the devices have been properly terminated.

 Verify that the cables have been properly connected. Also verify that
the cables are not defective.
In some cases the installation may actually have a driver problem as
opposed to a hardware problem.

Pinpointing the Problem

I remember setting up two servers in a cluster configuration, where both

servers were connected to a single external SCSI hard disk. A process that
should have taken a few hours turned into a few days because only one of
the servers could see the external disk. After verifying that the SCSI IDs
were set properly (several times over), not so carefully examining the
cables, and removing and reinstalling the drivers, I finally detected the
problem that had caused days of stress and headaches—a pinhole puncture
in the controller cable.

Upgrading IDE Hard Disks

Configuring IDE devices tends to be simpler then SCSI devices, but you are
more likely to be working with SCSI disks in a server environment. In any
case, you should still be familiar with the IDE upgrade process for the exam
and for on-the-job success. As with all hardware upgrades, you will use the
specific procedures outlined by the manufacturer. The following steps are
intended as a general outline of the things that need to be considered during
the upgrade process.
One of the most important considerations when upgrading IDE disks is
the master/slave relationship. If a single IDE disk is being upgraded, the new
disk will be configured as a master. If a second disk is being added to the sys-
tem for storage purposes, the original can be left as the master and the new
one designated as a slave. If the new disk will function as the main disk, then
it will be designated as the master and the existing one as the slave. Before

Copyright ©2002 SYBEX, Inc., Alameda, CA

Increase Memory 323

physically installing the disk, set the jumpers to establish the master/slave
relationship using the manufacturer’s documentation.

If you are installing a second drive from a different manufacturer, verify its
compatibility with the existing drive. Also, to save yourself some headaches,
make sure there is adequate cabling before actually beginning the upgrade.

Once the jumpers are set, the drive can be mounted into an empty bay and
the IDE cable attached to the new drive, making sure that pin 1 of the cable
is matched up with pin 1 on the drive. Once the drive has been installed, the
server can be restarted. Most servers will autodetect the drive, but you
should enter the system’s BIOS to ensure that the new drive is listed. The
drive should also be listed during the bootup process if it has been installed
correctly. Your final step will be to format and partition the drive; how you
do this will depend on the operating system installed.

Troubleshooting an IDE upgrade can usually be resolved by answering these
 Does the current BIOS support the size of the hard disk? (If not, a
BIOS upgrade will be necessary.)
 Has the master/slave relationship been properly configured?
 Is the cable properly connected? Has pin 1 on the cable been matched
to pin 1 of the hard disk?
 Is the power connected?

Increase Memory
Operating systems and software applications have RAM (random-
access memory) requirements that must be met before the software will run
properly. It seems as though each new software release needs more memory

Copyright ©2002 SYBEX, Inc., Alameda, CA

324 Chapter 10  Hardware Updates

than the one before. This is probably the main reason to upgrade the server’s
RAM. Fortunately, memory is one of the easier upgrades to perform. The
first things you want to check are the available space in the server to add
more RAM and the type of RAM currently installed. For example, if the sys-
tem supports up to 512MB of RAM and there is already 256MB installed,
obviously the maximum you can add is another 256MB. Before purchasing
RAM, answer the following questions:
 How much more RAM can the server support? This can quickly be
determined from the server’s documentation or from the manufac-
turer’s website.
 What type of RAM (SIMMs [single in-line memory modules]
or DIMMs [dual in-line memory modules]) is currently installed
in the server? (Do not mix EDO and non-EDO RAM or ECC and
non-ECC RAM.)
 What is the speed of the existing RAM? (The RAM that you add to
the computer must match the speed of the existing RAM.)
 What type of contacts are used? (DIMMs use gold for all contacts but
SIMMs can use tin or gold. Be sure the new RAM uses the same metal
as the existing RAM.)
Once you have determined the amount of RAM that can be added and
the specifications of the installed RAM, you are ready to make the purchase.
When purchasing new RAM, it is always recommended that you buy from
a reputable manufacturer. If you do decide to purchase off-brand RAM,
make sure you read the server’s documentation first to ensure that doing so
will not void your warranty. Keep in mind that some servers require the
RAM (usually SIMMs) be installed in pairs, but you can verify this through
the server’s documentation.

Make sure to check off-brand RAM for compatibility with your current system.

When you are ready to install RAM in the server, power off the server and
disconnect the power to the motherboard (you may need to disconnect a few
cables to get to the socket). Remove the existing RAM and install the new
memory module. It is fairly straightforward to do since the module can fit in
only one way.

Copyright ©2002 SYBEX, Inc., Alameda, CA

BIOS/Firmware Updates 325

After you perform the RAM upgrade, you may get an error message during
the POST informing you of a mismatch error. Don’t panic yet: Simply go
into the BIOS and verify that the new RAM is recognized, restart the server,
then save the changes and exit. This should clear the error message.

The following are some general things to consider when troubleshooting a
RAM upgrade:
 Is the RAM properly seated and inserted all the way into the socket?
 Try placing the RAM into a different socket.
 Does the server boot with the old RAM alone?
 Does the server boot with the new RAM?
 Does the new RAM meet all the requirements to co-exist with the
original RAM?
 Consider that the RAM may be faulty.

Once the RAM is successfully installed, server performance can be tuned

through the operating system or applications.

BIOS/Firmware Updates
Firmware (as discussed in Chapter 3, “Motherboards and Proces-
sors”) is software that controls hardware. Firmware updates on a server
include SCSI controllers, RAID controllers, tape drives, and CMOS BIOS.
Although other hardware devices have firmware, these four are the most
common firmware updates. When installing firmware updates be sure to
obtain the software from the product manufacturer. The potential risks of
component damage or inoperability should be a deterrent enough to not

Copyright ©2002 SYBEX, Inc., Alameda, CA

326 Chapter 10  Hardware Updates

download firmware updates from just any Internet site. Manufacturer’s sites
design and verify firmware updates specifically for their products. By down-
loading from the manufacturer’s site, you can reduce your risk.
One of the most common upgrades will be to the CMOS BIOS. Most
mainboards now use a flash ROM that can be reprogrammed countless
times using a flash utility. This means all you have to do is run an update util-
ity to upgrade the BIOS and the software will make all of the necessary mod-
ifications. You will need the make and model of the mainboard and the
revision number to locate the correct flash update and utility from the man-
ufacturer’s website. The download should contain data files, a flash utility,
and a Readme file. Flashing is not the only way to upgrade the BIOS but this
is the most common method with newer servers.

Make sure you download the correct BIOS update for your server. Flashing
the BIOS with the incorrect upgrade can leave your server unbootable.

A firmware upgrade can potentially leave your server unbootable so, like
any other upgrade you perform, a full system backup should be done before
proceeding. Also before upgrading the firmware, make sure to document
the current CMOS settings. Since some flash utilities clear the CMOS RAM,
you may need to restore some of your CMOS settings after the upgrade
is complete.

Applying a CMOS Upgrade

The following steps outline the general procedures for upgrading a system’s
BIOS. The exact procedures you follow will be guided by the manufacturer’s
documentation and may therefore vary slightly from those outlined below.
1. Download the upgrade from the manufacturer’s website. It’s a good
idea to download the current version as well if you don’t already have
it in case you need to go back to it. Some flash utilities will have an
option to back up the current BIOS. If the option is there, use it.
2. Most BIOS upgrades will include a Readme file. Review the contents
to determine how to upgrade the BIOS. Keep in mind that each man-
ufacturer will have its own procedures.
3. Create a bootable floppy disk (if required) and download the BIOS
update onto the floppy. In most cases it will be a zipped file and will
need to be decompressed to your floppy.

Copyright ©2002 SYBEX, Inc., Alameda, CA

BIOS/Firmware Updates 327

4. Restart the server using the floppy and start the upgrade as outlined in
the manufacturer’s instructions.
5. Once the upgrade is complete, remove the floppy and restart the
server. The new BIOS version should be displayed on the screen.
Proceed to the CMOS settings and reconfigure your parameters.

Again these steps are going to vary by manufacturer. Some may require the
CMOS settings to be cleared and others may require power to be removed
from the motherboard after the upgrade for a short period of time. This is why
it is important to carefully review the Readme file. If the upgrade is carried out
incorrectly, your server may be unbootable in the end.

Troubleshooting a CMOS Upgrade

If the upgrade was unsuccessful, you might hear one or more beeps as your
server starts up. Refer to the manufacturer’s website for error messages asso-
ciated with the beeps.
If the flash process was interrupted, your BIOS will only be partially
programmed or may now be corrupt. If this is the case you will need to con-
tact the manufacturer for a replacement BIOS chip (you may even be
replacing the entire motherboard).
If the BIOS update appeared to be successful but the system behaves errat-
ically and error messages appear, the wrong BIOS version has been installed
or it is corrupt. In either case, repeat the process to restore the original ver-
sion. If this restores the system to normal, verify that you downloaded the
correct BIOS for your server and repeat the upgrade.
If the system performs poorly after the new BIOS is installed, verify the
CMOS settings and make the necessary adjustments (you should have writ-
ten down the CMOS settings prior to performing the upgrade).
The motherboard is not the only hardware component with a BIOS. Other
components such as SCSI cards and tape drives have firmware that at some
point might need to be upgraded as well. Although there is a good chance an
upgrade of these components will never be necessary, you should still be
aware that they do exist. An upgrade might need to be applied to correct a
firmware bug, improve performance, or take advantage of new features.
Check the manufacturer’s website for the latest BIOS upgrades and proce-
dures on how to include them.

Copyright ©2002 SYBEX, Inc., Alameda, CA

328 Chapter 10  Hardware Updates

Replace the UPS

Depending on the hardware upgrade being performed, there might be
an impact on the system’s power consumption. For example, installing
redundant components within a server might require upgrading of the power
supply and/or the UPS. All servers should be connected to a UPS (uninter-
ruptible power supply). A UPS is a battery system that conditions incom-
ing power and maintains power to the server in the event of an electrical
failure. The battery is connected between the server and the electrical outlet.
When there is a power failure the battery system will run the server and any
other devices connected to the UPS for a period of time. Depending on the
size and strength of the battery system, a UPS may support minutes to hours
of power.
Software is available that will interface between the UPS and server allow-
ing for alerts and remote notification that the server is running off battery
power. Another key feature that UPS software provides is a shutdown option
for the server. If the electrical power is not re-established before the battery
runs out, the software will alert users, close all software programs that are
running, and then power down the server operating system. This prevents
the hard shutdown that occurs when power fails and all open programs
including the operating system are abruptly shut down. Hard shutdowns can
cause loss of data as well as program corruption. If the UPS that the server
is currently using is underpowered, then an upgrade to the UPS will also have
to be made. UPS software also monitors the power load being placed on the
UPS battery. If the load exceeds the recommended battery load, then the UPS
should be upgraded. Any time a hardware upgrade is performed, the UPS
load should be checked to confirm that it is capable of handling the stress of
the added component.
Working in conjunction with a UPS should be a surge protector. Some
manufacturers combine both a UPS and surge protector in one, but it is
advisable to purchase a dedicated surge protector. A surge protector has a
built-in circuit breaker that will trip should a spike of electricity arrive. This
prevents the momentary increase in electricity from damaging the computer.
Many surge protectors offer modem and network protection along with elec-
trical protection. This ensures that any potential damage traveling down a
telephone or network line will be caught before it can reach the computer.
All sensitive equipment should be protected by a surge protector (including
computers, printers, fax machines, and networking products).

Copyright ©2002 SYBEX, Inc., Alameda, CA

Upgrading Adapters 329

Upgrading Adapters
Upgrading adapters can include network interface cards, RAID con-
trollers, and SCSI cards. The upgrade may be in the form of a firmware
upgrade or replacing the old adapter with a new one. Either way, the pro-
cess of upgrading an adapter is fairly straightforward. If it is relatively
new hardware you are dealing with, you will probably be looking at a
software upgrade. If you are dealing with legacy hardware that is becoming
a bottleneck and performing poorly, chances are you will be looking at a
hardware upgrade.

Network Adapters
One of the most common adapters to be upgraded is the network adapter. In
most cases it is a fairly straightforward process, except of course when the
card is installed and doesn’t work. Then your troubleshooting skills once
again come into play.
When upgrading your network adapter, begin with a visual inspection of
the server and determine the type of slots available. Chances are you will be
using a PCI network card so you need a PCI slot available. You also want to
avoid resource conflicts, so determine what IRQs, I/O addresses, and mem-
ory addresses are available. Tools such as Microsoft’s Device Manager can
be used to determine what resources are available.
Install the network adapter (using ESD best practices) by removing the
metal plate if necessary and inserting the card. Once the card is seated all the
way into the slot, it should be secured using a screw. Not doing so may result
in the card creeping out of the slot and no longer working (or worse, causing
a short inside of the server). Once the NIC is installed and the server is reboo-
ted, you should verify that the link light on the card is lit.

Some NICs will come with diagnostic software that you can use to test that the
NIC is functioning correctly. The software can verify that the different compo-
nents on the NIC are functioning and can also provide diagnostic reports. The
software diagnostics can test network connectivity. If you don’t have diagnos-
tic software, one of the simplest tests to make sure the NIC is functioning is to
log onto the server from a client desktop or, if you are running TCP/IP as the
protocol, you can use the PING utility. If you are unsuccessful in pinging a
host, check the IP address, the speed the network card is set to, and verify that
it is not having a resource conflict with another device.

Copyright ©2002 SYBEX, Inc., Alameda, CA

330 Chapter 10  Hardware Updates

In this chapter you learned about some of the common hardware upgrades
that are often done to a server. CPUs, hard disks, and memory are the three most
common components to be upgraded in a server to improve performance.
All motherboards have limitations so, when upgrading a processor, con-
sult your documentation to determine the maximum speed supported by the
board. Also consider the design of the board—determine whether it is a slot
or socket design; this will impact on the type of chip your purchase. If you
are upgrading to multiple processors, keep in mind that the recommended
stepping between processors is one step.
When upgrading hard disks, begin by determining whether you will be
dealing with IDE or SCSI disks. If you are dealing with SCSI, you need to pay
attention to termination issues and the SCSI ID assigned to the new device.
With IDE disks you need to pay attention to the master/slave relationship.
Before going ahead and increasing the memory in a server, you need to
first assess the RAM that is currently installed. Consider how much memory
can be added and the type and the speed of RAM already present. The type
of RAM you choose should be supported by the manufacturer; some war-
ranties will be void if RAM from another manufacturer is used.
Firmware updates are applied to fix bugs and to take advantage of new
technologies. Two most important things to keep in mind are to download
the correct BIOS version for your system and to make sure not to interrupt the
flash process; interruptions can leave your system unbootable.
A UPS upgrade can involve upgrading the UPS battery, upgrading the UPS
software, or replacing the entire system with a new one.
Once a server has been installed and configured, it needs to be monitored
and maintained on a regular basis to ensure it continues to perform optimally
over time.

Exam Essentials
Know the general procedures to use when upgrading hardware. There
are some best practices (such as ESD) that should be adhered to when per-
forming the upgrade of any hardware component.
Know what to look for when upgrading a processor. Understand the
different things to check for when upgrading a processor or adding an
additional processor to a multiprocessor system.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Key Terms 331

Know how to upgrade a hard disk. Understand the difference in

upgrading an IDE disk as opposed to a SCSI disk. Each type of upgrade
has specific issues that you must consider.
Know how to perform a memory upgrade. Understand the general
steps involved in performing a RAM upgrade and what things to check
for before purchasing RAM for a server.
Know when to apply a BIOS update. BIOS updates can be applied to
fix known bugs or to take advantage of new technologies; not all revisions
need to be applied.
Know how to locate a BIOS update. BIOS updates can be located from
the manufacturer’s website. Make sure the BIOS applied is the correct
revision for your server.
Know how to apply a BIOS upgrade. Understand the general steps in
performing a flash upgrade.

Key Terms
B efore you take the exam, be certain you are familiar with the
following terms:

BIOS SCSI (small computer system

contacts SCSI ID
firmware slot
flash utility socket
hard shutdown stepping
IDE (intelligent drive electronics) surge protector
jumpers termination
master/slave relationship UPS (uninterruptible power
RAM (random-access memory)

Copyright ©2002 SYBEX, Inc., Alameda, CA

332 Chapter 10  Hardware Updates

Review Questions
1. It has become clear that your server needs a firmware update. You are
debating when to apply this update. Which is your best option?
A. During lunch when very few users are accessing the server.

B. During the middle of the workday.

C. The Friday night before you leave on vacation.

D. During the weekend.

2. A colleague comes to you, the lead technician, and says she noticed on
the hardware vendor’s website that there is a new firmware upgrade
available for your server. She thinks you ought to apply it. What will
you do? (Select all that apply.)
A. Apply the upgrade at once.

B. Check out the Readme that comes with the upgrade.

C. If your system is suffering symptoms, apply the update.

D. If your system is not suffering symptoms, keep the upgrade handy

but don’t apply it yet.

3. Name some limitations that you may encounter when considering

converting a server from a single processor to a multiprocessor system.
A. Lack of enough SCSI IDs

B. Lack of open IRQs

C. NOS limitations
D. Bus limitations

4. Which of the following are considerations when deciding whether

to implement multiprocessing? (Select all that apply.)
A. Will the OS support multiprocessing?

B. Will the motherboard support multiprocessing?

C. The maximum processor speed can’t exceed 1.2GHz.

D. The processor(s) must support symmetrical multiprocessing.

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 333

5. When applying a firmware upgrade, which of the following should

you do? (Select all that apply.)
A. Make sure you have all software and hardware components
necessary to perform the upgrade.
B. Download the upgrade.

C. Make sure you know how to reverse the procedure in case some-
thing goes wrong.
D. Read the Readme to find out what is involved with the upgrade.
E. Keep the server online throughout the entire procedure so users
have access to their data.

6. Monica is a network administrator who has been assigned the task of

upgrading the hard disks in an older server that has been running for
two years now. There is little documentation available for this server.
What two things must Monica immediately determine before she can
go forward with her hard disk replacement?
A. The number of disks in system

B. The serial number of the hard disks

C. The SCSI IDs of all the hard disks

D. The master/slave relationship

E. The type of hard disks that are in the computer

7. You’ve recently changed out your server’s IDE hard disk with a new
one but you can’t seem to get the hard disk to come up and be recog-
nized. There is an IDE CD-ROM in the system as well. What could be
the problem? (Select all that apply.)
A. BIOS doesn’t recognize the correct cylinders and heads.
B. CD-ROM is set to be master.

C. CD-ROM and hard disk are both set to be slave.

D. Termination jumper on hard disk isn’t set.

Copyright ©2002 SYBEX, Inc., Alameda, CA

334 Chapter 10  Hardware Updates

8. You need to upgrade the firmware in your server. What is the most
likely scenario for you to proceed with your upgrade?
A. Power the server off and upgrade the firmware.

B. Upgrade the firmware with the server operational.

C. Disconnect all users, but leave the server running to upgrade the
D. Power the server off, replace the motherboard, and then bring the
server back up.

9. What is likely to be your best source of information about the tasks the
firmware upgrade will accomplish and how long it will take?
A. The website at
B. The documentation that came with the server

C. The documentation that came with card

D. The documentation that came with the rack

E. The Readme file that came with the firmware upgrade

10. What are the hardware considerations when upgrading a computer to

a multiprocessor (SMP) system? (Select all that apply.)
A. The processor stepping.

B. The CPU clock speed.

C. The motherboard support for SMP system.

D. L2 cache on both the existing and new processor.

11. Which of the following should you consider when troubleshooting the
installation of a SCSI disk?
A. Master/slave relationship

C. Termination

D. Cables

Copyright ©2002 SYBEX, Inc., Alameda, CA

Review Questions 335

12. Bonnie is attempting to add some system memory to a server. The

system’s bus accepts 72-pin DIMMs. Bonnie is attempting to add one
new DIMM to the existing system. After she adds the new memory she
gets no errors, but the computer doesn’t recognize the new memory.
What could be causing the problem? (Choose all that apply.)
A. Incorrect speed on new DIMM.

B. Brand of DIMM isn’t compatible with rest of architecture.

C. Incorrect capacity of new DIMM.

D. Current DIMMs aren’t ECC.

13. Suzanne is working on a server that has four slots in it for DIMMs.
Two of the slots have 64MB DIMMs in them already. Suzanne wants
to add a 128MB DIMM, giving the system 256MB of total system
memory. When she adds the DIMMs, the power-on self-test memory
count shows the full 256MB but she now gets an error telling her to
adjust the BIOS. What could be the problem?
A. Nothing’s wrong.

B. Can’t pair DIMMs of different capacities.

C. First two DIMMs are ECC DIMMs, new one’s not.

D. First two DIMMs are silver-tipped, new one’s not.

14. You have a server that is RAM-starved. You purchase a DIMM from
a reputable memory manufacturing company, install it, and find that
the system won’t boot up. What could be the problem? (Choose all
that ap