Sharding in Practice

1. The Answer People Reach For Too Soon

Sharding means splitting one logical table across several databases, so that each machine holds only part of the data and serves only part of the traffic. It is the standard answer to "what do you do when one database is not enough," and it is reached for far earlier than it should be.

The gap between how sharding is discussed and how it behaves is wide.

• In interviews people reach for it far too soon. Candidates propose it within two minutes of hearing a large number, usually without naming a shard key or a query pattern.