In order to scale the Amazon database (RDS) you could either reboot the RDS instance to a larger size or create a slave RDS instance called a read replica. Adding more read replicas allowes for offloading of the read-only queries.
A nice new feature is that a read replica can now be promoted to a full standalone RDS instance. This will stop that instance being a read replica, so replication stops to it. It can now service write requests.
The benefits of promotion are :
1) Perform DDL operations:
Operations that take a long time, like rebuilding instances can be done on the read replica. Once completed and replication is up-to-date the instance can be promoted and web applications can now use the new database.
2) Sharding
Breaking a database up is a good way to split up tables that are not required for joins or duplicating tables across multiple instances.
Any read replica shards can then be promoted to standalone servers.
3) Recovery against failures:
Existing recovery methods were using the multi-AZ (availability zone) deployments or Point-in-time recovery. The new ability to promote a read replica can be used to quickly switch from a corrupt or offline master (source) RDS to a totally new one.
To promote a RDS – right click the current Master database and choose – “Promote Read Replica”.
Here is a link to the AWS blog post – http://aws.typepad.com/aws/2012/10/amazon-rds-for-mysql-promote-read-replica.html
