Amazon's two managed relational databases — distributed architecture, or the lower floor cost.
Pick Aurora for a MySQL- or PostgreSQL-compatible database that values its distributed storage: failover in roughly 30 seconds, up to 15 low-lag readers on one volume, a multi-region Global Database, and Serverless v2 scale-to-low. Pick standard RDS when you need an engine Aurora does not offer — MariaDB, Oracle, SQL Server, Db2 — or when a small, steady workload costs less on RDS's simple instance-plus-volume pricing. The single biggest factor is the engine: need MariaDB / Oracle / SQL Server and it must be RDS; otherwise it is Aurora's architecture against RDS's lower floor.
| Criterion | Aurora | RDS (standard engines) |
|---|---|---|
| Supported engines | MySQL- and PostgreSQL-compatible only | MySQL, PostgreSQL, MariaDB, Oracle, SQL Server, Db2 |
| Storage architecture | One distributed volume — 6 copies across 3 AZs | An EBS volume per instance; Multi-AZ adds a standby |
| Read replicas | Up to 15, sharing the volume, tens-of-ms lag | Up to 15 (MySQL/PostgreSQL), async, variable lag |
| Failover speed | Typically ~30 s when replicas exist | Multi-AZ instance: typically 1-2 minutes |
| Serverless / scale-to-low | Serverless v2 — 0.5-ACU steps, can pause to zero | No serverless option for standard engines |
| Multi-region | Global Database — up to 10 Regions, sub-second lag | Cross-Region read replicas, asynchronous |
| Pricing model | Compute + storage + I/O, or I/O-Optimized | Instance-hours + storage + provisioned IOPS |
| Cost for small steady workloads | Higher floor | Lower — simple instance plus a single volume |