Wednesday, September 14, 2016

Disaster Recovery reference architecture for data stores

It is always recommended to do DR planning at solution level which means that all the components making the solution should be well co-ordinated during DR. But to reach the solution level, it is important that DR design of individual components is ready.

This post documents a reference design of an important component of any solution which is Data/State store. This component is accessed by application tiers. This post is not about High Availability – Load Balancing combination. But, deals with DR scenarios where the secondary/DR site is in hot/warm/cold state.

Below diagram states the reference design of how the DR architecture commonly looks like. DR Agents play a critical role of when to call it a DR and based on that the app configuration changes to connect to available instance of Data Store.


So what are the considerations when designing the DR solution for Data Store ? Here are few important ones:

  • How application will store connection string to data store and how it will get updated at the time of DR.
  • How the DR site will be provisioned ?
  • What will be the state of DR site at normal times – hot/warm/cold. This will depend on RTO/RPO objectives and will also determine the provisioning method.
  • How the replication of data will happen between primary and DR site.
  • How the failover will happen from primary to DR site ?
  • How the failback will happen from DR to primary site ?

1 comment:

Shane Bond said...

Thank you. Cloud Disaster Recovery is really a nice concept - it is awesome, and I appreciate your helping to share about this.