Aurora

AWS-built, MySQL- and PostgreSQL-compatible relational engine. Same wire protocol and tools as RDS, but with a re-architected storage layer for higher throughput, durability, and elastic scaling.

What’s Different from RDS

• Distributed storage auto-grows to 128 TiB; data is kept as 6 copies across 3 AZs, self-healing.
• Storage is decoupled from compute — replicas share one storage volume, so adding readers is cheap.
• Up to 15 read replicas with a reader endpoint that load-balances reads.
• Failover typically < 30 s (promote an existing replica).

Deployment Options

• Provisioned — pick instance classes (like RDS).
• Serverless v2 — fine-grained autoscaling in ACUs; scales with load, good for variable/spiky workloads.
• Global Database — one primary region + up to 5 read-only secondary regions, < 1 s cross-region replication; cross-region DR with fast promotion.

Handy Features

• Backtrack (MySQL) — rewind the cluster to a point in time without a restore.
• Fast cloning — copy-on-write clones for test/analytics without duplicating storage.
• Endpoints — cluster (writer), reader (load-balanced reads), and custom endpoints for subsets.

When to Choose

• You want relational + joins + transactions, but RDS read scaling, failover speed, or storage ceiling is limiting.
• Variable load → Serverless v2. Multi-region reads/DR → Global Database.
• Pure key-value at extreme scale → DynamoDB instead.