The previous chapter focused a bit on optimization in terms of splitting the query across cloned instances, which would be only part of a true database scalability concern. The database would still have performance concerns, as there is replication lag coupled with a fragile splitting mechanism through the proxy. The best route to take to reduce all of these concerns, is to architect a distributed database from the very beginning.

In the sharding write pattern, we take the previous concepts a bit further in-depth, by not analyzing the query to determine which instance to execute against. Instead, we use a cluster management tool called MySQL Fabric, which was announced by Oracle in early 2014. Fabric provides a single API to create and manage farms (data centers) of databases or even farms of farms of databases. Using this method, we are able to create multiple database instances, and from a Fabric node, group them into many different layouts to achieve high availability...