Implementing VTL and DeDuplication Technology

1
Implementing VTL and De-Duplication Technology

• Gavin Cole
• Storage Consultant SEE
• Sun Microsystems

1
2
Explosive data growth impacts performance and
ability to maintain Back-up SLA'sKeep
everything, delete nothingBackup/Restore
failures causing increased riskManagement
complexity making back-up process
ineffectiveCost to store data estimated at
5,935/TB per year
Factors impacting your ability to protect data

• Explosive data growth impacts performance and
  ability to maintain Back-up SLA's
• Keep everything, delete nothing
• Backup/Restore failures causing increased risk
• Management complexity making back-up process
  ineffective
• Cost to store data estimated at 5,935/TB per year

Back-up Window Data Volume Value of Data Storage
Complexity Budgets
Source ESG Study
3
Market Trends
- InfoPro 2008
4
Backup Storage Optimization
5
Storage Optimization Cost SavingsThree Methods

• Straight disk for small TB configurations
• VTL with hardware compression
• Best used for performance sensitive back-ups
• Half the physical storage space needed to store
  your back-up data (21 Compression ratio)?
• Compression is applied across all storage within
  VTL
• Data De-duplication
• Flexibility Policy based
• Data De-duplication price best /GB for optimized
  capacity
• Look for duplicity in data AND long retention
  timeframes
• Best used for disaster recovery replication and
  storing months of back-up versions
• Engineering performance and/or higher
  utilization of storage

6
Virtual Tape Libraries
Open Systems Tape Emulation Improves Performance
and Enhances Backup

• Backup application thinks its talking to a tape
  library
• Actually backing up to VTL disk buffer
• Data can the be de-duplicated, replicated and/or
  migrated to tape
• Many more virtual tape resources
• Disk performance with tape look, feel, and cost
  structure
• Data movement without impact to servers or backup
  window

• Emulated
• Library
• Drive
• Cartridges

Tape Automation
Sun VTLAppliance
7
Do I need a VTL ?
Size of file systems used for D2D
Global amount of data
Restore times
Size of the DB
Number of backup policies
Number of HBA and OS
CPU usage
Primary/Secondary Disk Array
Number of volumes to be vaulted
Drive Sharing Module
Backup Server
SAN Disk
Multiple backup applications
SAN Throughput (difficult to feed)?
Cost of drive sharing module
Vault
Incremental Backup on tape take time to restore
Media Server
Number of tape drive to manage
Slow IP network
Cloning times
Number of media server
SAN Tape
Start/Stop Mechanism
Hardware error
IP Network
Tape drive Throughput (difficult to feed)?
Type of client (Oracle,SAP,Lotus ..)?
8
Integrated Tape with your VTL
9
Integrated Tape VTLAutomated Tape Caching

• The VTL acts as cache to the physical tape
  library, providing transparent access to data
  regardless of its location.
• Tapes always appear to be inside virtual
  libraries
• Backup application will always have direct access
  to data
• VTL maps directly to tape resources
• Policy based migration of virtual cartridges to
  physical tape
• Time based - daily or weekly
• Policy based (can be combined)
• Age of data
• Disk space used
• End of backup
• Provides flexible space reclamation policies
• Free space immediately upon migration,
• After specified retention period,
• When running out of space

VTL Tape Library
10
Integrated VTL NDMP 4.0

• NDMP 4.0 Using NetBackup Direct to Tape copy
• Backup application is in complete control
• Mount requests - physical drive and tape
• What data is copied
• VTL is only the data mover
• Symantecs Veritas NetBackup 6.5.1 and above only
• VTL Plus only - requires NDMP software option

11
Integrated VTL NDMP 4.0 with ACSLS
VTL
BackupClients
NetBackup 6.5 Master and Media servers
SAN
data mover
mount request
ACSLS
12
Integrated VTL ISV Copy Process
BackupClients
Tape Automation
MIAV(Man In A Van)
Backup servers
S A N
VTL Virtual Tape Library

• ISV backup application manages virtual and
  physical tape devices and cartridges
• Backup application is in complete control
• Mount requests - physical drive and tape
• What data is copied
• Highest Performance

13
De-Duplication
14
What is data de-duplication?

• A means through which software analyzes data
  streams, using a hashing algorithm, to determine
  repetitive data
• A high-performance database is maintained to
  store single instances of each unique block of
  data
• If data is identified as identical to existing
  data in the database, only a pointer to that data
  is stored, not a new instance of data
• To restore data, pointers must be matched to the
  data in database, and data must be reconstituted

15
Data de-duplication Weighing the benefits
Premise
Considerations

• RELIABILITY Because only one instance of data is
  stored, a single corruption will impact many
  references
• COMPLEXITY Adds several additional steps to
  existing backup processes
• PLANNING Actual compression ratios very
  unpredictable
• PERFORMANCE Extremely CPU-intensive, causing
  performance degradation
• COST Required high-performance disk server may
  negate cost savings of de-duplication

• COST Significant budget space savings are
  possible because only one instance of data is
  stored
• COST Double-digit compression ratios can be
  achieved
• COST Bandwidth savings for replication
• Can be significant

16
Without Data De-Duplication

Sample parameters Data volume 20TB 5
growth and weekly change Onsite Retention 5
weeks
110TB
86TB
63TB
41TB
20TB
Week 1
Week 5
Week 4
Week 3
Week 2
17
With Data De-Duplication
Sample parameters Data volume 20TB 5
growth and weekly change Onsite Retention 5
weeks
86 reduction
Data Stored 15.2 TB Redundant data NOT stored
94.8TB
20TB
15.2TB
14TB
12TB
10TB
6.6TB
Week 1
Week 5
Week 4
Week 3
Week 2
18
Ultra-efficient replication
19
Data Replication VTL Prime

• VTVs data blocks are Hashed (de-duped) resulting
  in Virtual Index Tapes (VITs)?
• VITs are replicated based on standard VTL IP
  replication polices to the remote VTL Prime (VTL
  to VTL)?
• The remote VTL Prime PULLS new data blocks
  (De-dup. repository to De-dup. repository)?
• Only net new blocks are retrieved/pulled
• Global data de-duplication

20
Suns VTL Prime ConfigurationSpoke and Hub
21
Never Forget Tape
22

0kWh 0CO2
Data at Rest On Tape
23
Off-Site Protection
24
Thank you!

• Gavin Cole
• gavin.cole_at_Sun.com

24