Overview
What is a Network?
Two or more computers are connected together by a medium and are sharing resources. These resources can be files, printers, harddrives, or CPU number-crunching power.
A network can consist of two computers connected together on a desk, or it can consist of many Local Area Networks (LANs) connected together to form a Wide Area Network (WAN) across a continent.
Many individuals have asked to see the "Big Picture" of networking: How does everything . Where does Microsoft NT fit in with routers and the OSI layers? What about UNIX, Linux and Novell? The big picture in the following slide attempts to show all areas of networking and how they tie into each other.
Storm Clouds Telecommunications media or Information Providers that connect to the Internet
Machine symbol Network "linker" can be a bridge, router, brouter or gateway Jagged haphazard dotted line - the Internet
ISO/OSI Model
The International Standards Organization (ISO) Open Systems Interconnect (OSI) is a standard set of rules describing the transfer of data between each layer in a network operating system. Each layer has a specific function. For example, the physical layer deals with the electrical and cable specifications. The OSI Model clearly defines the interfaces between each layer. This allows different network operating systems and protocols to work together by having each manufacturer adhere to the standard interfaces. The application of the ISO OSI model has allowed the modern multiprotocol networks that exist today.
7. 6. 5. 4. 3. 2. 1.
Application Layer (Top Layer) Presentation Layer Session Layer Transport Layer Network Layer Data Link Layer Physical Layer (Bottom Layer)
ISO/OSI Model
The OSI model provides the basic rules that allow multi protocol networks to operate. Understanding the OSI model is instrumental in understanding how the many different protocols fit into the networking jigsaw puzzle.
Local Loops
LANs
MANs
WANs
Voice Lines Modem Connections 56 kbps ISDN (Integrated Services Digital Network) - 2 x 64 kbps digital lines ADSL (Asymmetrical Digital Subscriber Line) - up to 8 Mbps * Cable Modems - up to 30 Mbps
Cable modems are not part of the local loop but do fall into the category of the last mile, or how high speed digital communication gets to the premises (home). It would incredibly expensive to replace the existing cabling structure. And because this cabling was designed for voice communications rather than digital, all of these protocols are needed to overcome the existing cabling limitations in the local loop and provide high speed digital data transmission.
Cabling standards
Hardware Protocols
Cat 3, 4 and 5 cables IBM Type 1-9 cabling standards EIA568A and 568B Ethernet cabling standards: IEEE 802.3 (10Base5), IEEE 802.3a (10Base2), IEEE 802.3i (10BaseT) Unshielded Twisted Pair (UTP) Shielded Twisted Pair (STP) Connectors: RJ45, RJ11, Hermaphroditic connectors, RS-232, DB-25, BNC, TEE
Network Interface Cards (NICs) Repeaters Ethernet Hubs or multi port repeaters Token Ring Multi Station Access Units (MSAUs), Control Access Units (CAUs) and Lobe Access Modules (LAMs) Bridges
Ethernet frame types: Ethernet_II, Ethernet_SNAP, Ethernet_802.2, Ethernet_802.3 Media Access Control layer (MAC layer) Token Ring: IBM and IEEE 802.5 Logical Link Control Layer (LLC) IEEE 802.2 TCP/IP IPX/SPX Asynchronous Transfer Mode (ATM)
RS232, V35
References
1. Introduction to Networking and Data Communications Eugene Blanchard Edited by Joshua Drake, Bill Randolph and Phuong Ma 2. Computer Networking: A Top-Down Approach Featuring the
Internet
Jim Kurose & Keith Ross 3. Internetworking Technology Overview Cisco Systems 4. Internetworking Case Studies Cisco Systems
Network Topology
Overview of Network Topology and Case Study of Flat Neighborhoods
Examples of well known designs follow this slide, we shall assume all topologies are using 100 Mbit Ethernet as the medium and rate them on design categories.
Bus Topology
Bus Topology
Ring Topology
Ring Topology
Star Topology
Star Topology
Robustness Very Good Efficiency Very Good Simplicity Poor Scalability Excellent
A New Topology is Born In the past, it has been standard to come up with a topology first, and then adapt it to certain tasks. Modern design philosophy has changed this practice. Now a subset of problems or needs gives rise to special task network designs. One such design has been conceived right here at UK.
Brought about by the need to build a large cluster supercomputer from common networking components. Driven to evolve from the need for (more) efficient communication between cluster nodes.
Multiple small, interleaved subnets link each machine by a number of one-switch latency paths. Any machine can belong to as many subnets as it has network cards onboard. Sounds simple, but several problems arise from the design.
The wiring scheme and subnets can now be designed by a piece of software developed in the KAOS lab. This problem appears to be NP Complete (Very Bad) and must be solved using a genetic search algorithm. A simplified version allows you to design your own FNN on the web. http://aggregate.org/FNN/
Throw out bottom 2/3 results and replace with mutations thereof. Merge Subnets of pairs in top 1/3 results. Re-Evaluate and rate accordingly
Each machine in the cluster swaps unique identifiers with all of its neighbors at boot up. Address resolution is done locally using the table that this swap generates. Non-Dynamic Solution
Assembled on April 11, 2000 in the KAOS lab by Dr. Dietz and Mr. Mattox Fully Functional on April 16 The first working implementation of an FNN
KLAT2
Total Value: $41,205 Peak Performance: 64 GFlops $643.83 / GF Total Value: $1.5M / yr Peak Performance: 672 GFlops $2,232.14 / GF / yr
Superdome
Advantage
KLAT2
KLAT2
Purchase new Nodes Upgrade the Old Nodes Recompute Scheme Rewire EVERYTHING
Purchase a new Cabinet Plug and Play
Superdome
Advantage
Superdome
The Lowdown
Number if NICs in each node PCI Bus Speed Increased Physical Distance Complexity of Design
Use of KLAT2
KLAT2 is mainly a lab experiment, thus its practical uses are limited :
Insufficient Non-Volatile Storage Weak Back-Up System Slow Internet Connection to the WAN Limited Application Compatability
With further R+D, the FNN cluster may evetually bring about a supercomputer in every home movement.
Summary
Topology Development Philosophy has Evolved Special Purpose Topologies use Networks to Solve Specific Problems Network Topologies are Always Expanding
The Credits
http://www.engr.uky.edu/ece/faculty/dietz/index.html
http://aggregate.org/KAOS/ http://aggregate.org/FNN/ http://www.ccs.uky.edu/~douglas/ http://www.ccs.uky.edu/~connolly/ http://sdx.uky.edu/
Topics:
Why IP?
What is IP?
Protocol suit defining an interface between lower level hardware functionality and higher level application oriented protocols. Provides a least common denominator for all network hardware. Provides best effort service for datagram delivery from host to host.
How?
How?
Fields
Version(4 bits) 4 Header Length(4 bits) Size of the header in 4 byte words. Type of Service(8 bits) Mostly unused. Length(16 bits) Total length of IP datagram in bytes.
Fields continued
Identification(16 bits) unique identifier Flags(3 bits) 0, Dont fragment, More fragments. Fragment Offset(13 bits) Offset of fragment in 8 byte words.
Time To Live (8 bits) Hop count. Protocol(8 bits) Higher level protocol address. Header Checksum Ones compliment sum of all 16 bit words in IP header.
Fields, more?
Source Address(32 bits) Where it came from. Destination Address(32 bits) Ummm, you know.
Fragmentation
IP only requires ~500 byte MTU from hardware layer but allows for packet sizes up to 65535 bytes. IP datagrams can be fragmented into smaller packets to travel over various networks then reassembled at the destination.
Fragmentation
Fragments from the same datagram carry the same identifier field. All fragments except the last have the More Fragments bit set. The Offset Field is an index into the original datagram payload.
IP Addressing
Hierarchical (cuz thats what CS people do) 32 Bits long. Globally unique (most of the time.) Assigned to network adapter, not host. Composed of network part and host part. Hosts on the same physical network have the same network address.
IP Addressing
Class A - [0][7 Bit Network][24 Bit Host] Class B - [10][14 Bit Network][16 Bit Host] Class C - [110][21 Bit Network][8 Bit Host]
IP Addressing
Classless IP addressing (the way it really is.) Arbitrarily long network portion followed by host portion. Can not tell dividing line from IP address. A netmask is used to divide the address.
IP Forwarding
Each host has a table with tuples of network addresses, address length, next hop information, and interface information. To forward an IP packet, find the longest network address that matches destination address. Send the packet out the corresponding interface to the next hop (may be local.)
IP Forwarding
Example:
Interface0 = 128.163.125.2/24 Interface1 = 24.249.125.187/24
Address/Length 128.163.125.0/24
Interface Interface0
128.168.0.0/16
24.249.125.0/24 0.0.0.0/0
128.163.125.1
Local 24.249.125.1
Interface0
Interface1 Interface1
Whats Next?
IPv6 128 bit addressing (more people can play quake.) Fewer fields for simplicity
Overview
Mobility in the Internet Basic Mobile IP Protocol IMHP : Route Optimization in Mobile IP Other Issues
May change point of network connection frequently May be in use as point of network connection changes Usually have less powerful CPU, less memory and disk space Less secure physically Limited battery power
Mobile computers are one of the fastest growing segments of the PC market Short-range wireless networks (Bluetooth) available from IBM, Toshiba, Dell, HP High-speed (11 Mbps) wireless LAN products are now easily and cheaply available (IEEE 802.11a, IEEE 802.11b) Low speed (currently 128 Kbps) Metropolitan Area Wireless Network services are available in some cities and spreading (Metricoms Ricochet)
Problem with current IP .It assumes that a nodes IP address uniquely identifies its point of attachment to the Internet Mobility alternatives without Mobile IP .On moving, change IP address Use host-specific routes(using LSR) to reach mobile hosts .Mobility vs. Portability
Functional Entities in Mobile IP : -Mobile Node -Home Agent -Foreign Agent Each mobile node is assigned a unique home address within its home network When away from home network, it is assigned a care-of address either by : -Registering with a Foreign Agent -Obtaining a temporary IP address
Basic Mobile IP
H.A.
Correspondent node
F.A.
M.H.
Protocol Overview
Agent Discovery
Extension of ICMP Router Discovery protocol Used by mobile nodes to discover Foreign Agents and to detect movement from one subnet to another Mobility Agents (H.A.s and F.A.s) periodically broadcast agent advertisements
Mobile node expects to receive periodic advertisements If it doesnt receive them, it deduces that either -it has moved OR -its agent has failed Mobile node can also broadcast Agent Solicitation messages
Registration
Mechanism by which M.H. communicates reachability info to its H.A. Registration messages create or modify a mobility binding at a H.A., which is then valid for a certain lifetime period Uses 2 control messages sent over UDP -Registration Request -Registration Reply
Replay Protection : Needed to ensure that registration messages are not replayed by a malicious host. Done using : -Nonces OR -Timestamps
Registration Authentication
Concern : Forged registrations permit malicious hosts to remotely redirect packets destined for the mobile host Default authentication between M.H. and H.A. uses MD-5 with a shared secret key No authentication between M.H. and F.A.
Delivering Datagrams :
When the mobile host is away, H.A. intercepts packets addressed to the M.H. and tunnels them to the M.H.s care-of address The tunneling scheme could use either of : - IP-in-IP Encapsulation -Minimal Encapsulation
Broadcast Datagrams -A H.A. forwards a broadcast datagram only if the M.H. requested forwarding of broadcast datagrams (in the registration request) Multicast Datagrams -M.H. can use a local multicast router -M.H. can use a bidirectional tunnel to its H.A.
IMHP
Extension to the basic Mobile IP protocol that features : -Route Optimization -Authentication of Management packets Defines four entities : -Mobile Hosts -Local Agents -Cache Agents -Home Agents
Triangle Routing in basic Mobile IP -Limits performance transparency -Creates bottleneck at Home Agent
H.A.
Correspondent Node
F.A. M.H.
Route Optimization
Eliminates triangle routing Any correspondent node can maintain a binding cache Correspondent node tunnels datagrams directly to the care-off address of the Correspondent mobile host Node
M.H.
F.A.
H.A.
Binding Management
Four message types : -Binding Warning -Binding Request -Binding Update -Binding Acknowledge Lazy notifications are used (except MH to HA and previous FA)
As part of registration, M.H. requests its new F.A. to notify its previous F.A. New F.A. sends binding update to prev F.A. Previous F.A. updates its binding cache entry for the M.H. and sends a binding ack. Authentication of binding update is based on a shared registration key
Special Tunnels
When a F.A. receives a tunneled datagram for a M.H. for which it has no entry, it is tunneled back to the H.A. in a special tunnel Gives the datagram one more chance of successful delivery Avoids possible routing loops
Authentication in IMHP
IMHP has simple authentication procedures which preserve the level of security in todays Internet is defined to make use of strong authentication
M.H. to H.A. authentication -strong authentication based on a shared secret General Authentication -a random number specified in binding request is echoed in the reply by the H.A.