Replication

Replication keeps copies of a database on more than one machine, with changes to the primary copied continuously to the replicas.

It buys two different things. Availability: if the primary fails, a replica can take over, so a disk failure is minutes of disruption rather than a restore from backup. And read capacity: reporting and search queries can be pointed at a replica so they do not compete with the traffic customers are waiting on.

The catch is replication lag. A replica is a fraction of a second behind, which is invisible for a report and very visible if a user saves a change and is then shown a stale read. Anything that must reflect a write immediately has to read from the primary.