Before we can really decide how to spread our data across several database servers, we need to find appropriate candidates. To do this, we should start at the database level for databases that are extremely active. What qualifies as extremely active? Databases that fit any of these criteria are a good start:
The database experiences more than 10 million transactions per day
The database handles more than 100 million queries per day
The database writes more than 100 million tuples per day
Once we've chosen a database for horizontal scalability, we need to look at its tables and decide which should be distributed. Tables that make good choices are those that fit one or more of the following criteria:
Tables that contain more than 10 million rows
Tables that experience more than 1 million writes per day
Tables that are larger than 10 GB
This recipe will discuss easy ways to find prospective tables for further study.