Estimate replication lag from write volume and apply throughput.
Replication lag is a queueing problem: if writes arrive faster than the replica applies them, lag grows without bound rather than settling. Single-threaded apply is the usual constraint. Reading from a lagging replica returns stale data, which is why utilisation above 80 per cent is a correctness concern rather than just a performance one.
Database Replication Lag
Lag grows whenever the write generation rate exceeds the replica apply rate
Lag grows whenever the write generation rate exceeds the replica apply rate Replication lag is a queueing problem: if writes arrive faster than the replica applies them, lag grows without bound rather than settling. Single-threaded apply is the usual constraint.
Reading from a lagging replica returns stale data, which is why utilisation above 80 per cent is a correctness concern rather than just a performance one.
This calculator takes 5 inputs: Writes per second on the primary, Bytes per write, Replica apply throughput, Network latency to the replica, Existing backlog. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.