Disk Based Disaster Recovery

1
Disk Based Disaster Recovery Data Replication
Solutions

• Gavin Cole
• Storage Consultant SEE

2
Agenda

• Planning for a Disaster
• Using Local Copies for Protection
• Using Remote Mirroring for Protection
• Conclusion

3
Agenda

• Planning for a Disaster
• Using Local Copies for Protection
• Using Remote Mirroring for Protection
• Conclusion

4
Disaster
any interruption in the normal access to a valid
set of data used by applications and end users to
execute mission critical business processes for
an unacceptable period of time.
Jon William Toigo Chairman, Data Management
Institute 2005
5
Source Ontrack Data Report 2007
6
Source Ontrack Data Report 2007
7
The threat of data loss cant be ignored

• Risks
• High cost of data loss and downtime
• Insufficient data recovery plans and procedures
• 70 of businesses fail after major data loss
• Recovery overextends limited staffs and financial
  resources
• Needs
• High availability and fast data recovery
• Seamless integration with existing IT
  infrastructures
• Interoperability with current and future storage
  and computing systems

8
The Cost of Downtime

• Forrester Consulting Interviewed 138 companies
• Estimated cost of 1 hour downtime
• Less than 10,000 / hr 25
• 10,000 to 100,000 / hr 33
• 100,000 to 500,000 25
• 500,000 to 1 million 13
• Greater than 1 million 4
• 67 - could not estimate the financial cost of
  downtime

9
Key definitions
Resumption of normal business
Last safe Backup
Disaster
Recovery Point Objective (RPO)? The time between
the last safe backup and the point of time of the
disaster
Recovery Time Objective (RTO)? The time elapsed
from when the disaster occurred to the resumption
of normal business activities
10
Business Continuity

• Planning to never go down
• Always have access to information
• Needs more than a good data recovery strategy
• A disaster can be something as simple as a
  deleted file
• Use disk duplication strategies
• Mirroring
• Snapshot
• Remote replication

11
Complete data protection requires personalized
and practical solutions
Risk
Disruption
Regional Disaster
Seconds
Local Disaster
Minutes
Business goals Threats Budget realities Existing
assets
Hours
Operator error
Days
Hardware failure
Application
Department
Data Center
Enterprise
Scale
12
Seven Key Planning Steps

• Business impact assessment
• How long can I live without data?
• Discovery
• What data do I need first?
• Budget
• What is my data worth?
• Role-based teams
• Who are the key people?
• Data protection
• How do I protect what I need?
• Logistics
• What are the physical requirements?
• Testing
• Will my plan work?

13
Agenda

• Planning for a Disaster
• Using Local Copies for Protection
• Using Remote Mirroring for Protection
• Conclusion

14
Volume Copy Terms

• Complete point in time replication of one Volume
  (source) to another (target) within a Storage
  Subsystem

Copy Pair
Target or Copy or Clone
Target Volume that maintains a copy of the data
from the source
Source Volume that accepts host I/O and stores
application data
15
How Volume Copy Works
16
Snapshot Terms

• A point-in-time (PiT) image of a volume
• Logical equivalent of a physical copy

Storage System
Physical Disk Space
Base Volume - the volume from which the Snapshot
will be created
17
Snapshot Flow Chart
18
Using Volume Copy and Snapshot together

• Copying the Snapshot creates a full PiT clone
  copy while I/O continues to base volume (LUN)?

19
Agenda

• Planning for a Disaster
• Using Local Copies for Protection
• Using Remote Mirroring for Protection
• Conclusion

20
Remote Volume Mirroring

• Ongoing, real-time replication of a volume from
  one storage system to another

21
Remote Volume Mirroring Components

• Primary volume accepts read and write host I/O
• Secondary volume accepts read host I/O.
• accepts remote writes of data from controller
  owner of Primary volume.

22
Remote Volume Mirroring Components

• Mirror Repository volume Stores mirroring data,
  such as info about remote writes that have not
  completed

Mirror Repositories
Mirror Repositories
Mirror Pairs V1 -gt V1M V2 -gt V2M V3 -gt V3M
Primary
Secondary
Primary
Secondary
23
Synchronous Replication

• The primary disk system acknowledges a host write
  when the data has been successfully mirrored

• Primary benefit
• Ensures remote data is an exact replicate of the
  local data
• Note only effective for campus area replication

24
Asynchronous Write Mode

• Allows the primary disk system to acknowledge a
  host write request before the data has been
  successfully mirrored

• Primary benefits
• Reduces impact of latency when replicating over
  longer distances
• Provides performance improvement compared to
  synchronous for primary site I/O (disk system
  and application)?
• Enables effective replication over longer
  distances (WAN)?

25
Preserved Write Order

• Write operations to the secondary disk system
  matches I/O completion order on the local disk
  system
• Also referred to as a consistency group
• Primary benefit
• Maintains data integrity in multi-LUN
  applications (databases) by eliminating
  out-of-order updates at the remote side that can
  cause logical corruption

26
Remote Volume Mirroring Mirror Management

• Role Reversal (from secondary to primary or vice
  versa) is user-initiated
• If primary is also base volume for snapshots,
  role reversal will cause associated snapshots to
  fail
• It is possible to force role change for the local
  volume if communication to the remote volume is
  down
• Used in disaster recovery scenarios
• Can prepare by mapping secondary volumes to hosts
  using Storage Partitions before they are promoted

27
DR / HA Architecture
Cluster 1
Cluster 2
Site A
Site B
Volume 3 replication
Volume 1 2 replication
M mirror
M mirror
28
Architecture Description

• Site A and Site B each contain a copy of
  critical data
• Critical data is copied in real time using the
  disk controllers
• minimal impact on server processing power
• OS and key applications are clustered across
  both sites
• If either site fails application transparently
  fails over to remote
• Customers notices minimal disruption
• 2 way Disaster Protection
• Cluster 1 uses volume V1 and V2 primary
  business is at Site A, Mirrored to Site B for
  protection
• Cluster 2 uses volume V3 - primary business is at
  Site B, Mirrored to Site A for protection
• Sites are connected by Fibre Channel network for
  performance
• Could be connected by long distance IP network
  will be performance impact on replication.

29
Agenda

• Planning for a Disaster
• Using Local Copies for Protection
• Using Remote Mirroring for Protection
• Conclusion

30
Could you survive a disaster?

• 35 of companies have a plan
• 60 of plans are never tested
• Half the companies that suffer losses never
  recover
• 10 50 K per MB to re-create data

31
Seven Key Planning Steps

• Business impact assessment
• How long can I live without data?
• Discovery
• What data do I need first?
• Budget
• What is my data worth?
• Role-based teams
• Who are the key people?
• Data protection
• How do I protect what I need?
• Logistics
• What are the physical requirements?
• Testing
• Will my plan work?

32
Thank You
Disaster Recovery Planning

• Gavin Cole
• gavin.cole_at_sun.com
• 33 6 70 72 99 53