Overview
Amazon Aurora is the AWS-engineered relational database, providing MySQL- and PostgreSQL-compatible engines on top of a custom distributed storage layer that decouples compute from storage and replicates writes synchronously across six storage nodes spanning three Availability Zones. Aurora delivers throughput of roughly 3–5x stock MySQL and 2–3x stock PostgreSQL on equivalent hardware, with sub-second failover and continuous backup to S3.
The current Aurora portfolio spans Aurora Standard (provisioned), Aurora I/O-Optimized (capped I/O cost for write-heavy workloads), Aurora Serverless v2 (granular auto-scaling at 0.5 ACU increments down to zero), Aurora Global Database (cross-Region replication with sub-second lag), and Aurora DSQL — the new active-active distributed SQL service that reached general availability in 2025 and competes with Spanner and CockroachDB. Aurora is the recommended target for most enterprise MySQL and PostgreSQL workloads on AWS.
Key Features
- MySQL 8.0-compatible and PostgreSQL 14–17-compatible engines
- Distributed storage replicating 6 copies across 3 Availability Zones
- Aurora Serverless v2 with 0.5 ACU granular scaling and scale-to-zero
- Up to 15 low-latency read replicas with sub-100ms replica lag
- Aurora Global Database for cross-Region replication and disaster recovery
- Aurora DSQL — multi-Region active-active distributed SQL (Postgres-compatible, 2025)
- Aurora Limitless Database for sharded write scaling on PostgreSQL workloads
- Babelfish for Aurora PostgreSQL — wire compatibility with T-SQL applications
- Zero-ETL integration with Redshift, OpenSearch, and SageMaker Lakehouse
- Aurora Machine Learning for in-database calls to Bedrock and SageMaker
- Performance Insights, Database Activity Streams, and DevOps Guru integration
- Backtrack (MySQL) for rewind without restoring from backup
Pricing
| Configuration | Model | Cost (us-east-1) |
|---|---|---|
| Aurora Serverless v2 | Per ACU/hour | $0.12/ACU/hour (Standard); $0.16/ACU/hour (I/O-Optimized) |
| db.r6g.large (provisioned) | Per instance/hour | ~$0.29/hour Standard, ~$0.38/hour I/O-Optimized |
| db.r6g.4xlarge | Per instance/hour | ~$2.32/hour Standard |
| Storage (Standard) | Per GB/month | $0.10/GB/month + I/O charges |
| Storage (I/O-Optimized) | Per GB/month | $0.225/GB/month, no I/O charges |
| Aurora Global Database (cross-Region) | Per replicated I/O | $0.20 per million replicated I/Os |
| Aurora DSQL | Per DPU + storage | ~$8/DPU/hour, $0.81/GB/month |
Pricing verified May 2026 from AWS public pricing for us-east-1. I/O-Optimized typically reduces total cost by 30–55% for write-heavy workloads versus Standard once I/O charges are factored in. Reserved Instances offer up to 65% discount on provisioned instances.
Strengths
- Significantly higher throughput than stock MySQL/PostgreSQL on equivalent hardware
- Continuous backup with point-in-time recovery to the second, no separate maintenance
- Aurora Serverless v2 truly scales to zero, eliminating cost for development environments
- Global Database with sub-second cross-Region replication is rare among managed services
- Aurora DSQL provides a strongly-consistent distributed SQL option without leaving AWS
- I/O-Optimized billing simplifies cost forecasting for write-heavy workloads
Limitations
- Material lock-in — Aurora-specific extensions and DSQL are not portable to vanilla engines
- I/O charges on Standard can become the dominant cost line for write-heavy workloads
- Aurora Serverless v2 scale-from-zero has 5–15 second cold-start latency
- Versioning trails community PostgreSQL by 6–9 months on new releases
- Cross-Region read latency is non-zero — DSQL or Global Database has architectural trade-offs