RDS Multi-AZ vs Read Replicas: Availability vs Read Scaling
4 min read · Updated 2026-10-25
Multi-AZ and read replicas both create copies of your RDS database, but for opposite reasons. Confusing them is one of the most common exam mistakes — and AWS knows it.
Multi-AZ — for availability
Multi-AZ maintains a synchronous standby replica in a different AZ. The standby serves no traffic — it exists only to take over automatically during failover (typically 60–120 seconds, no connection-string change needed). It's a disaster-resilience feature, not a scaling feature.
Read replicas — for read scaling
Read replicas are asynchronous copies that serve read-only traffic. Applications must explicitly point read queries at the replica endpoint. Replicas can also be promoted to standalone primaries — useful for DR or migrations — and can live in other Regions for cross-Region reads.
Memory aid
- 'Automatic failover / minimize downtime' → Multi-AZ
- 'Offload read traffic / reporting queries / cross-Region reads' → read replica
- Need both? They're not exclusive — a primary can be Multi-AZ AND have replicas
- Aurora behaves differently: replicas there are multi-purpose (both HA and reads)