jargon

Comparison

Leader-follower replicationvsMulti-leader replication

Leader-follower replication

all writes go to one node and the other copies exist only to be read from and to be promoted if it dies.

The standard topology: one node accepts writes and streams its change log to read-only followers. It avoids write conflicts entirely because there is only one writer. Write throughput is capped by that one node, and every failover is a moment where you must decide between waiting and losing data.

writescopiesClientClientLeaderFollowerFollower
Full entry →

Multi-leader replication

two regions both accept writes to the same record and something now has to decide which version wins.

A topology where more than one node accepts writes and they replicate to each other. It buys local write latency and survives a region loss without promotion. It introduces write conflicts as a permanent design problem, and last-write-wins conflict resolution silently discards data.

writecopyClientClientLeaderLeaderConflictingversions
Full entry →

Related comparisons