Anda di halaman 1dari 25

Avamar Technical

Review
Q4-2013
New Hire SE

Copyright 2010 EMC Corporation. All rights reserved.

Module1:
Avamar
Fundamentals
Copyright 2010 EMC Corporation. All rights reserved.

EMC Next Generation Data Protection


Deduplication Backup Solution
End-to-end, software/hardware solution
Integrated system for simple, predictable results
Client-side deduplication; within & across clients

Avamar
VM

Improves backup window, less network load


Backup process minimizes data sent and stored
Reduces network and virtual infrastructure stress

Integrated high availability and reliability


RAIN for high availability and fault tolerance
Recoverability verified daily
Full backups, every time one-step
recovery
DR via replication
Higher backup success rate &
reliability
Increased ROI, lower TCO, less risk
Copyright 2010 EMC Corporation. All rights reserved.

Avamar Fundamentals
Avamar shrinks backup window, speeds data recovery
TRADITIONAL BACKUP

Long backup cycles


Laborious restore process

Copyright 2010 EMC Corporation. All rights reserved.

AVAMAR BACKUP

Up to 10x times faster backups


Simple, one-step recovery

Avamar Fundamentals
Avamar daily full backups versus traditional daily full
backups
Data Type

Amount of
Primary Data
Backed Up

Amount of
Data Moved
Daily

Reduction
Factor

Windows file systems

3,573 GB

6.1 GB

585:1

Mix of Windows, Linux, and UNIX file systems

5,097 GB

11.7 GB

435:1

Engineering files on NAS (NDMP backups)

3,265 GB

24.2 GB

134:1

Mix of 20% databases, 80% file systems


(Windows and UNIX)

9,583 GB

80.0 GB

119:1

Mix of Linux file systems and databases

7,831 GB

104.2 GB

75:1

While results will vary by data type and mix, Avamar


can
dramatically improve backup performance and
Source: EMC
efficiency
Copyright 2010 EMC Corporation. All rights reserved.

Avamar Fundamentals
Unlike anything else in the industry

Reduce:
Data moved by up to 99.7%
Network interface card usage at the
client by 99%
Backup windows by 90%
Client CPU usage by 80%
Disk access at the client by 50%
Memory usage at the client by 50%

Copyright 2010 EMC Corporation. All rights reserved.

Avamar Fundamentals
Building Block - Avamar Nodes
Scale Up

Spare Node

Up to 16
Data Nodes

Multi Node
RAIN
System
11-124TB
De-dupe
Capacity

Single Node System


Utility Node
1.3-7.8TB De-dupe Capacity

Copyright 2010 EMC Corporation. All rights reserved.

*1TB Primary Data, 80/20 File System to Database, 90days FS, 30 days DB

Avamar Fundamentals
Inside a 7.8TB storage node

1 x Quad core CPU


32GB RAM
12x 2TB 7.2k SATA drives
RAID1 Layout
100GB SSD
4 x GigE Ports (8 in Gen4S)
Dual Power Supplies
2 Rack Units or 2U

Copyright 2010 EMC Corporation. All rights reserved.

Front

Back

Avamar Fundamentals
Avamar fault tolerance for reliable protection and
access
Avamar server
Verified
checkpoi
Layers of Systematic Fault
nt
Tolerance
RAID protection from disk failures
Parity
across
storage
nodes

Utility and
spare node

Redundant Array of Independent


Nodes (RAIN) protection from
storage node failures
Checkpoints (snapshots) for
Operational Failures
Grid-to-Grid Replication for site
failures and disasters
Daily data integrity and
recoverability checks

Copyright 2010 EMC Corporation. All rights reserved.

Walking the File System


3
8k 25k
RAW
DATA 9k 6k 17k
k

2 4
k k

5
k

13k

3
k

9k

12k

6k

2. Compression Reduces chunks by 30-50%

SHA-1 Algorithm

3. Initial Hash Cycle creates Atomic Hashes


20 byte

20 byte

20 byte

20 byte

20 byte

20 byte

20 byte

20 byte
20 byte
20 byte

20 byte
20 byte
20 byte
20 byte

20 byte

20 byte

20 byte
20 byte

SHA-1 Processes the data and creates a unique


fixed length 20-byte Data String.
4. Sticky-byte Factoring cycle creates Composite
Hashes.
5. Composite Hash is rehashed to new unique 20
byte hash
6. Sticky-Byte Factoring and Hashing continues
until one unique Root Hash is obtained for the
file

20 byte

Root Hash

1. Initial Snapup: Data is separated into chunks


Chunks Vary in size between 1 byte and 64K
bytes.
Data chunks average 24K bytes in size.

Copyright 2010 EMC Corporation. All rights reserved.

10

Checking Data for Backup


3
8k 25k
RAW
DATA 9k 6k 17k
k

12k

Is present?

N
Is atomic hash?

9k

6k
20 byte

3
k

20 byte

20 byte

20 byte

20 byte

N
Store hash
Send next hash

Y
Store Hash
Send & Store Chunk

20 byte
20 byte
20 byte
20 byte

20 byte
20 byte
20 byte

20 byte

20 byte
20 byte
20 byte
20 byte

Root Hash

5
k

20 byte

13k

20 byte

2 4
k k

Discard Hash
Snapup Complete

Local Client
Local Hash Cache
Local File Modified Date Cache

Copyright 2010 EMC Corporation. All rights reserved.

11

Avamar Fundamentals
How it Works ( Backup Data Structure )

Copyright 2010 EMC Corporation. All rights reserved.

12

Avamar Fundamentals
How it Works ( Backup Tree Index )

Initial
back up

Copyright 2010 EMC Corporation. All rights reserved.

back up 1

back up 2

13

Avamar Fundamentals
Avamar manages multi-sites from a single location
Intuitive, web-based
interface
At-a-glance dashboards
Capacity reporting and
alerting
Specialized
management interface
for desktop and laptop
clients

Copyright 2010 EMC Corporation. All rights reserved.

14

Avamar Fundamentals
Client Manager
Manage
Move, Retire, Delete or
Modify

Analyze
Backup, Restore, Failures

Upgrade or Downgrade
Remote Push

Activate
Search and activate by OU

Copyright 2010 EMC Corporation. All rights reserved.

15

Avamar Fundamentals
High-speed, scalable Avamar backups to Data Domain
systems
Broadens Avamar use cases, and
solves more customer problems
Supports specific data types: Oracle;
Microsoft SQL Server, Exchange, SAP*,
Sybase*, SharePoint; and VMware
images

Direct backups to optimal systems


based on workload attributes, not
technology
Data Domain integration provides
Combines
simplicity
access toAvamar
an additional
540and
TB of
efficiency
with Data Domains scale
usable capacity
and performance
Copyright 2010 EMC Corporation. All rights reserved.

16

Flexible Deployment Options


Avamar Data Store options meet your needs
Avamar Data Store
Single Node

Avamar Virtual Edition


Virtual Appliance

Remote Offices

VMware Environments

Sized for distributed office workloads


Centralize via replication (required)
Fast backup, one-step, local recovery

Avamar server, in a virtual machine


Guest and (VADP) image backups
Fast, changed-block recovery

Avamar Data Store


Scalable RAIN Grids

Avamar Business Edition

Enterprise Environments

Mid-Market Environments

Scalable nodes
RAIN and high availability
Up to 124 TB

Copyright 2010 EMC Corporation. All rights reserved.

AVE

Single node, sized for small data centers


Optional replication for DR/centralization
Lowest $/TB Avamar solution

17

Module 1: Knowledge Review


Smallest number of Nodes needed for an Avamar RAIN Grid?
3
Name the 4 Layers of Avamar Systematic Fault Tolerance:

Checkpoint / HFScheck
RAID
RAIN
Replication

Average Avamar Chuck Size ?


24K (compressed to ~12K) (what happens when source data is already compressed
like a .zip or .gz file ??)
List some issues with 1 Single Node (including attached to Single DD) w/No Replication ?
Fault Resiliency, can only handle Disk Drive Failure
Average Avamar change rate per day (backups) ?
and thus for the year/52weeks/365 days ?
How Much will a customer save in Capacity if system sized for 45 dailies, and now reduced to 30 days

No. of DD systems that can be integrated behind


One Avamar ?
Up to 5 have been tested

Copyright 2010 EMC Corporation. All rights reserved.

18

Module Summary
Avamar as a Next Generation Backup Solution
Truly a Technology Differentiator by creating a
Full Backup every time
Significant reduction in the amount of Time
and Resources Needed to do a Backup
Ability to add capacity (storage) along with
processing as needed in the customers
environment
Integration w/other BRS, like DPA, Networker
and Data Domain
Avamar is first Data Protection solution that
incorporates Software and Hardware

Copyright 2010 EMC Corporation. All rights reserved.

20

Avamar and
the Clock

Copyright 2010 EMC Corporation. All rights reserved.

21

Backup/Blackout/Maintenance Windows
Standard Maintenance Activities used to be cron jobs. Starting in Avamar 5.x
the server manages maintenance activities internally within user-configurable
Windows

in
o
P

ps
u
ck
a
B

ck
e
Ch C
G
&

Copyright 2010 EMC Corporation. All rights reserved.

t
S
HF

ck
e
Ch

ps
u
ck
a
B

22

Avamar Default Clock Settings


New Operation

Backup
Full operation
Max 72 backups per node
Maintenance
HFS Check
Garbage collect
Max 20 backups per node
System maintenance

12:00am

6:00am

6:00pm

Backup

Maintenance

Backup

Copyright 2010 EMC Corporation. All rights reserved.

23

Backup/Blackout/Maintenance Windows
Standard Maintenance Activities used to be cron jobs. Starting in Avamar 5.x
the server manages maintenance activities internally within user-configurable
Windows

ps
u
ck
a
B

Copyright 2010 EMC Corporation. All rights reserved.

nt
i
ck
Po
e
k
ec
Ch
h
S
C C
HF
G
&

ps
u
ck
a
B

24

The Ideal 24 Hour Avamar


Schedule
Daily Schedule (Default ) starts at 6 a.m. and must always
complete before 8 p.m. (less than 14 hours)
Includes 3 hours of garbage collect
Allocate up to 8 hours for hfscheck

Replication starts at 10 p.m. and must complete by 6 a.m. (8


hours)
Replication starts after 80% of the backups have completed (2
hours)

All backups must start after 8 p.m. and must complete by 6


a.m. (up to 10 hours)

Exception is a few small clients can back up between 12 noon


and 3 p.m.

.m.

9 a.m.

12 noon

3 p.m.

CP
GC

6 p.m.

9 p.m.

12 midnight

3 a.m.

6 a.m.

CP
Hfscheck
Ltd Backups

Copyright 2010 EMC Corporation. All rights reserved.

Replication
Backups

25

THANK YOU

Copyright 2010 EMC Corporation. All rights reserved.

26