Page 1
© 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.© 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.
Disaster Recovery Site on AWS:
Minimal Cost Maximum EfficiencyRyan Holland, AWS
July 10, 2014
Page 2
What You Will Learn
• Disaster Recovery and Business Continuity
• Why AWS for disaster recovery?
• Common DR architectures
• Backup and restore
• Pilot light
• Warm Standby
• Hot Standby
• Customer case study
• Where to go next
Page 3
Disruptions to Business Continuity
Caused by outage of IT infrastructure
Affects businesses of all kinds and sizes
Can be very expensive
Page 4
Downtime
Natural Disaster
Security Incidence
Equipment Failure
Human Error
What causes downtime
Page 6
Conventional Disaster Recovery Sites
• High cost
• Low ROI
• Implemented only for most critical systems
• Usually scaled down to 50% of production
• Systems in a remote region challenging
• Costly software licenses based on hardware usage
Page 7
Disaster Recovery on AWS
• Unprecedented capabilities to implement DR sites
• Easily set up DR sites on different geographic regions
• Cut down DR site cost by up to 70%
• Substantial savings on software licenses
Page 8
Global Reach from Your Desktop
Page 9
Tools for Implementing DR on AWS
• Leverage tools like CloudForamtion to automate deployment.
• Choose an AMI strategy that fits the RTO requirements.
• Cross-region EBS snapshot and AMI copy
• Cross-region read replicas for Amazon RDS for MySQL
• Amazon Route53 and Auto Scaling
• EC2 reserved instances
Page 10
AWS Storage Options
Simple Storage ServiceHighly scalable object storage
1 byte to 5TB in size
99.999999999% durability
Elastic Block StoreHigh performance block storage device
1GB to 1TB in size
Mount as drives to instances with
snapshot/cloning functionalities
GlacierLong term object archive
Extremely low cost per gigabyte
99.999999999% durability
Page 11
Common DR architectures
Each architecture
differs from the other
In terms of RTO, RPO and Cost
Page 12
Simple to get started
Easy starting point for exploring the AWS cloud
Low technical barrier to entry
Focus on incorporating cloud into your DR
strategy, not on complex technical issues related
to hot-hot systems
Lowest cost
Very high levels of data durability at low price
Cost of storing snapshots in Amazon S3
Archiving possibilities beyond tape using
Amazon Glacier
Backup & Restore Architecture
Page 13
Back up and restore
Create instances from AMIs
Restore datafrom backups
Page 14
Many Ways to Back Up
Page 15
Backup & Restore Considerations
• Make sure you keep your AMIs current
• Use CloudFormation or other automation tools
• Consider EC2 light utilization reserved instances
• Test your DR plan frequently. Then test some more.
Page 16
Build resources around
replicated dataset
Keep ‘pilot light’ on by replicating core
databases
Build AWS resources around dataset and
leave in stopped state
Scale resources in AWS in
response to a DR event
Start up pool of resources in AWS when
events dictate
Scale up the database instance to handle
production capacity
Pilot Light Architecture
Page 17
Pilot Light Architecture
Page 18
Create instances from
AMIs
Pilot Light Architecture
Page 19
Activating a Pilot Light DR Site
• Use CloudFormation and Auto-Scaling to stage infrastructure.
• Keep your AMIs or bootstrapping scripts current.
• Leverage EC2 heavy utilization reserved instances for the
database
• Test your DR plan frequently. Then test some more.
Page 20
Build a environment similar to
production at a reduced scale
Keep data and files synchronized between
production and DR site by replication
Use smaller and/or fewer instances than Production.
Scale resources in AWS in
response to a DR event
Scale out the environment by adding more
instance
Scale up the instances to handle production
capacity
Warm Standby Architecture
Page 21
Warm Standby Architecture
Page 22
Warm Standby Architecture
Page 23
Moving Warm Standby to Production
• Use CloudFormation and Auto-Scaling to resize infrastructure.
• Leverage EC2 heavy utilization reserved instances for the
database and the warm standby instances.
• Test your DR plan frequently. Then test some more.
Page 24
Build DR site as mirror image of Production
Keep all data and files synchronized between production and DR site by
synchronous replication if possible
Pick the size and number of instances based on acceptable level of performance
without any change in case of a DR event.
Use RI (Reserved Instances) for capacity reservation and cost saving
Multi-site Architecture
Page 25
Load balance between
production and DR
If latency and error propagation risk
between production and DR sites are
acceptable
Multi-site Architecture
If DR site is isolated then
Switch over to AWS
Make necessary DNS changes to
redirect traffic to the DR site on AWS
Page 26
Multi-site Architecture
Page 27
DR site on AWS can be for
• Primary site on customer data center
• Primary on AWS itself
Page 28
Primary and DR Sites on AWS
Page 36
What enabled this?
• Eight isolated S3 regions
• AWS CloudFormation allows quick bootstrap of
another region.
• Route 53 latency based routing and failover
Page 37
User in San
Francisco
eu-west-1 (Ireland)
us-east-1 (Northern Virginia)
us-west-1 (Northern California)us-west-1 (Northern California)
DNS Failover
Page 38
What didn’t go wrong
• Official NYC evacuation map stayed up
• USA TODAY Weather map stayed up
• Thousands of other maps used for weather
reporting, data visualization and coordination
around the event all stayed up
Page 40
© 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.© 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.
Disaster Recovery Site on AWS:
Minimal Cost Maximum EfficiencyRyan Holland, AWS
Thank you!