Chapter 31
Cloud Database Comparison
Cloud Database Comparison: RDS vs Aurora vs PolarDB
Cloud databases eliminate operational overhead but differ significantly in architecture and performance. Choosing wrong is costly.
Product Landscape
| Product | Vendor | Architecture | MySQL Compat |
|---|---|---|---|
| RDS for MySQL | AWS | Traditional M/S, EBS storage | Full |
| Aurora MySQL | AWS | Disaggregated storage, 6-copy | Partial |
| PolarDB MySQL | Alibaba | Disaggregated, shared storage | High |
| Cloud SQL | GCP | Traditional M/S | Full |
Aurora vs RDS
Aurora rewrites the storage layer: data replicated across 6 copies in 3 AZs. Writes only need 4/6 quorum, reads 3/6. Redo logs go directly to storage, bypassing double-write buffer — 3-5x write throughput (real-world 1.5-3x).
Key Comparison
| Dimension | Aurora MySQL | PolarDB MySQL |
|---|---|---|
| Read replica lag | <100ms | <1ms |
| HTAP support | No | Yes (IMCI) |
| Global DB | Aurora Global DB | PolarDB Global DB |
| Ecosystem | Global best | China best |
Selection Guide
- AWS ecosystem / global: Aurora MySQL
- Spiky traffic SaaS: Aurora Serverless v2
- China market OLTP: PolarDB for MySQL
- HTAP (no ETL): PolarDB + IMCI
- Zero code change migration: RDS MySQL