Managed database services eliminate DBA overhead — backups, patching, replication, and failover are handled by the provider. The trade-off is less control over configuration and, in some cases, more vendor lock-in.
Relational (SQL) Databases
| Provider | Service | Engines |
|---|---|---|
| AWS | RDS | MySQL, PostgreSQL, MariaDB, Oracle, SQL Server |
| AWS | Aurora | MySQL/PostgreSQL-compatible, 5x MySQL performance |
| Azure | Azure SQL | SQL Server (fully managed), + MySQL, PostgreSQL, MariaDB |
| GCP | Cloud SQL | MySQL, PostgreSQL, SQL Server |
| GCP | Spanner | Globally distributed relational — horizontal scale |
| OCI | Autonomous Database | Oracle, self-tuning and self-patching |
Aurora and Spanner stand out for scale: Aurora separates compute from storage (scale each independently), and Spanner offers globally consistent transactions across regions — a capability no other managed service matches.
Oracle Cloud’s Autonomous Database is the most automated relational option, handling query tuning, index management, and patching without DBA intervention.
NoSQL Databases
| Provider | Service | Type |
|---|---|---|
| AWS | DynamoDB | Key-value / document; single-digit ms at any scale |
| AWS | DocumentDB | MongoDB-compatible document store |
| Azure | Cosmos DB | Multi-model: document, key-value, column, graph |
| GCP | Firestore | Document store for web/mobile |
| GCP | Bigtable | Wide-column; HBase-compatible |
DynamoDB is the most battle-tested at scale. Cosmos DB is the most flexible — one service that speaks multiple API dialects (MongoDB, Cassandra, Gremlin, SQL).
In-Memory / Cache
- AWS ElastiCache — managed Redis and Memcached
- AWS MemoryDB — Redis-compatible with durable multi-AZ storage
- Azure Cache for Redis — managed Redis
- GCP Memorystore — managed Redis and Memcached
Redis is the standard choice for session caching, rate limiting, pub/sub, and leaderboards.
Data Warehouse
- AWS Redshift — columnar, petabyte-scale, integrates with S3 data lake
- GCP BigQuery — serverless; pay per query, no cluster management
- Azure Synapse — combined data warehouse and big data analytics
BigQuery is often cited as the most operationally simple — no cluster to provision, scale, or tune.
Decision Guide
- Existing relational workload, no special requirements → Cloud SQL / RDS (PostgreSQL or MySQL)
- Need extreme scale with relational guarantees → Aurora or Spanner
- Oracle-licensed → OCI Autonomous Database
- High-throughput key-value at web scale → DynamoDB
- Multi-model flexibility → Cosmos DB
- Serverless analytics → BigQuery