General Considerations for BackupRestore Design with TSM

1
General Considerations for Backup/Restore Design
with TSM
TSM benefits

• Jason Bamford

Pan-EMEA Tivoli Storage Specialist
Seminarium BACKUP RESTORE w srodowisku
operacyjnym TSM 12.09.2002 Prezentuje Andrzej
Dominik CK ZETO S.A. TSS Manager
2
Contents Importance of
design Typical hardware architectures Defi
nitions of What does performance mean to you
? Things to consider In summary
3
Importance of Design
TSM is very versatile Small
environments Very small systems entry
level Could operate on a single system with
one tape drive Functionality can be limited by
hardware resources Important to understand what
can be expected through design
4
TSM is very versatile Mid range Top
performer Highly flexible design
possibilities Wide client support Extensive
storage device support Powerful features like
incremental forever Detailed design will
exploit the best functionality of the product
5
TSM is very versatile High End World class
storage management tools Central point of
control for a distributed enterprise Able to
handle multiple policies simultaneously Full
media tracking tools Complex projects simply not
possible without good design
6
Typical Hardware Architectures
7
(No Transcript)
8
Hardware Architectures (2)
Multiple Single Servers
Multi Single Server (2) Dedicated service for a
single application server. Can be managed
centrally. Scalable performance, but high
management overhead. Applicable to distributed
site with limited bandwidth
9
Hardware Architectures (3)
Multiple Single Servers/Tape Sharing
Multi Single Server/Tape Library Sharing
(3) Maximize value from tape library Central
media management Single points of failure
Application Server
Backup Server
Application Server
Backup Server
Application Server
Backup Server
TSM Server
TSM Server
TSM Server
Tivoli Data Protection for Oracle
Tivoli Data Protection for Oracle
Tivoli Data Protection for Oracle
Data
Data
Data
10
Hardware Architectures (4)
Centralised Backup Server
Central Backup Server (4) Central point of
management Simply to manage LAN bandwidth
hungry Single server could become a bottleneck
LAN
11
Hardware Architectures (5)
Centralised and Distributed Backup Servers
TSM Server
WAN
Tape Library
12
Hardware Architectures (6)
LAN
LAN Free Data Transfer
Application Server
Backup Server
LAN Free Data Transfer (6) Exploit SAN high data
transfer bandwidth TSM server will not
bottleneck with data traffic Scalable SAN
infrastructure costs relatively high
Tivoli Data Protection for Exchange/SAP
TSM Server
SAN
Data
Tape Library
13
Hardware Architectures Summary
Single Server (1) Dedicated service for a
single application server. Entry level solution
for small environments. Scalable performance,
but high management overhead. Multi Single
Server (2) Dedicated service for a single
application server. Can be managed
centrally. Scalable performance, but high
management overhead. Applicable to distributed
site with limited bandwidth Multi Single
Server/Tape Library Sharing (3) Maximize value
from tape library Central media
management Single points of failure
14
Hardware Architectures Summary
Central Backup Server (4) Central point of
management Simply to manage LAN bandwidth
hungry Single server could become a
bottleneck Wide Area Distribution (5) Central
point of management and configuration over wide
area Electronic media vaulting
potential Potentially WAN bandwidth hungry LAN
Free Data Transfer (6) Exploit SAN high data
transfer bandwidth TSM server will not
bottleneck with data traffic Scalable SAN
infrastructure costs relatively high
15
Definitions Of ..
16
What is a . Backup ?
Copying data from client workstation to server
storage to ensure against loss of data. The
server retains copies of multiple versions of a
file according to policy. Policy includes the
number of version and the retention time for
versions. (source TSM Admin Guide GC35-0410-00,
page 6)
17
What is an . Archive ?
Copying data from client workstation to server
storage for long-term storage. The server
retains archive copies according to policy for
retention time. (source TSM Admin Guide
GC35-0410-00, page 6)
18
What is . Space Management ?
Freeing up client storage space by copying files
from workstations with TSM to server storage.
This process is also called hierarchical storage
management (HSM). On the client TSM replaces the
original file with a stub file that points to the
original in server storage. (source TSM Admin
Guide GC35-0410-00, page 6)
19
What Backup is not good for
Keeping out of cycle copies of workstation data
for long term storage. Backup is designed to
keep versions of data based on its existences or
deletion from a workstation.
20
PerformanceYes please, Id like lots
But what does Performance mean to you ?
21
I see performance as .

• I have a small backup window, so need fast
  backup
• I need rapid recovery from disaster, fast
  restore is more important
• I restore or retrieve files regularly. Time to
  data is most important
• I manage large archives. Space efficiency/cost
  is most important
• Disaster recovery issues. I need off-site
  copies as quickly as possible
• I do space management. Rapid recall of files
  in needed (near line)
• We have a large community of laptop users with
  infrequent LAN access

Using Policy design Storage pool design Hardware
design A single TSM server could address all
these performance requirements
22
...I have a small backup window, so need fast
backup...

• Fast data transfer
• Minimum or no waiting for tape mounts
• Dont transfer files unnecessary

- Dedicated LAN or SAN - Stage files in disk
storage pools - Incremental forever
Dedicated or high speed LAN
Disk Storage Pool
Offline task migrate to tape
Source data
23
...I need rapid recovery from disaster, fast
restore is more important...
- Dedicated LAN or SAN - Recent staged data /
collocated media - Place clients data in
separate tapes

• Fast data transfer
• Minimum or no waiting for tape mounts
• Reduce media contention

Dedicated or high speed LAN
Collocation avoids contention for media allow
more parellel operations
Disk Storage Pool
Restore area 1
Restore area 2
24
...I restore or retrieve files regularly. Access
time is important...

• Identify files needing fast response
• Use dedicated stg pools (disk tape)
• Retain data on disk by age/size
• Consider type of tape technology

- Create separate mgmt classes for them - Dont
waste disk stg on files not needing - Large file
direct to tape - Mid-point loading systems have
better average seek times
Domain Policy
Collocation on
Workstation Source Volume
Backup Files lt 5Mb
Management classes
Backup Files gt 5Mb
Copy groups
Archive Files lt 10Mb
Archive files Backup files
Archive Files gt 10Mb
25
...I manage large archives. Space efficiency/cost
is most important...

• Reduced cartridge count
• Maximize cartridge utilization
• Minimize server reclamation/overhead

- Used very high capacity media - Do not
collocate data - Keep Archives of different
types in Separate storage pools
Collocation off
Domain Policy
Workstation 1
Management classes
As long and short term files will not be on same
media. Short term files expiring will not trigger
long term files on same volumes to be reclaimed
time and time again. Reduced reclaim overhead
Archive 5 years
Workstation 2
Copy groups
Archive 3 months
3 month retain 5 year retain
26
...We have a large community of laptop users with
infrequent LAN access...

• Need efficient dial-up protocol
• Method of shipping full data-set out
• Security for internet and dial-up users

- Adaptive diffencing technology - Instant
archive, rapid restore - Secure protocols
27
...We do space management. Rapid recall of files
in needed (near line)...

• Analyze data by size
• Maximise data availability
• Consider tape technology choice

- Use disk pools to reduce media wait - Keep
migrated files away from backups - Mid-point
loading media will give you quicker average
time to data
Domain Policy
Workstation 1 Volume using migration
Migration Stg pool files lt 20MB highmig100
Management classes
files gt 20MB highmig80 lowmig50 cacheon
Copy groups
Collocation on
Migrated files only stg
First disk stg - gives best recall
response Second disk stg - gives best migration
response Tape stg - gives you capacity you require
28
...Disaster recovery issues. I need off-site
copies as quickly as possible...

• Use pools in a remote library
• Keep database backup off-site
• Protect against media loss

- Exploit SANs to give speed and distance -
Backup directly to a remote library - Create
copy pools back at site A
TSM Primary
Copy Library
Primary pools Library
TSM Standby
Site A
Site B
Distance to 10Km
SAN Fabric
SAN Fabric
29
Things to consider
30
Know the characteristics of your data.
Domain Policy
Finance System Development Area
Disk Stg pool
Management classes
All files pool
Copy groups
Single management class All data stored in small
pool, and retained with small characteristics.
But some of these types have different needs
Single policy little control limited scope to
tune or change
31
Know the characteristics of your data.
Various storage needs of the data Finance System
Test area - - - - Only really need latest
version Database - - - - Core component
needs many version Spool print - - - -
Never get extra versions, re-creatable Temp
files - - - - Short useful life, can be big,
exclude Development Area Test database - - -
- Get deleted and recreate from live
regularly why back it up if you
can recreate it Spool print - - - - Only
for test purpose, poorly managed, exclude
Temp - - - - Short useful life, can be big,
exclude Source code - - - - Important to
keep many versions for developers. typically
lots of small files. Higher number of
restore requests generated from this area
Finance System Test area Database
Binaries Spool print Temp
files Development Area Test database Spool
print Temp Source code
Domain Policy
Management classes
Database

• Better understanding more efficient use of
  storage
• All data is not the same, so dont treat it
  that way.
• Take control of YOUR data, alter your polices
  for different data types
• Consider the impact of storing one data type
  with another in the same pool
• Tuning disk storage pools could vastly improve
  restore requests hitting disk not tape

Copy groups
Disk Stg pool
Binaries
Max file size 1MB
Source
Drill down into the applications
32
Know the characteristics of your data.
Storage Requirements for Source code Regular
small restores - - - - Benefit from
disk pool 20,000 files (70 lt 50k) - - - -
A disk storage pool with max file size
of 50k could satisfy 70 of restore
requests without mounting a tape
Disk pool size - - - - 7 versions
14,000 files x 7 versions x 50k
14,000 x 7 x 50 4.9GB
33
In Summary
34

• The Messages to Remember TSM is Flexible
• Can be optimised to many environments
  through policy storage pool design - even on
  one server
• Has solutions from low bandwidth to SAN
• Design IS important - big small
• Take control - understand your data at a low
  level