Managed Databases

Checking access...

Managed databases free you from the operational burden of installation, patching, backup, and replication. Every major cloud provider offers a portfolio of database services spanning relational, key-value, document, and caching engines.

Relational Databases (SQL)

Amazon RDS

RDS supports six database engines: MySQL, PostgreSQL, MariaDB, Oracle, SQL Server, and Amazon Aurora (a MySQL/PostgreSQL-compatible engine).

Features:

Multi-AZ: Synchronous standby replica in another AZ for high availability
Read replicas: Up to 15 asynchronous replicas for read-heavy workloads
Automated backups: Daily snapshot + 5-minute transaction log retention (configurable up to 35 days)
Storage auto scaling: Increases storage when free space drops below a threshold

Aurora separates compute and storage. The storage layer is a distributed, self-healing cluster that replicates data across three AZs with 6 copies. It offers 5x the throughput of standard MySQL and 3x of standard PostgreSQL.

Info

Choose Aurora when you need MySQL/PostgreSQL compatibility with higher performance and availability than standard RDS. Use standard RDS when you need engine-specific features (Oracle extensions, SQL Server integration) not available in Aurora.

Azure SQL Database

Fully managed SQL Server with built-in high availability, automatic tuning, and serverless compute options. Azure SQL Database offers:

Elastic pools: Share resources across databases
Hyperscale tier: Up to 100 TB per database with fast scaling
Geo-replication: Readable secondaries in paired regions

Cloud SQL (GCP)

Managed MySQL, PostgreSQL, and SQL Server. Features include:

Automatic storage increase: No downtime
Read replicas: Cross-region support
Private IP: Connect via VPC peering, no public internet exposure

NoSQL Databases

Amazon DynamoDB

Fully managed key-value and document database with single-digit millisecond performance at any scale.

Key concepts:

Partition key + sort key: Determines data distribution across partitions
On-demand vs. provisioned capacity: On-demand for unpredictable workloads, provisioned for stable traffic
DAX (DynamoDB Accelerator): In-memory cache for microsecond response times
Global tables: Multi-region, multi-master replication

Tip

Use DynamoDB for high-traffic applications with simple access patterns (lookups by primary key). If you need complex joins, aggregations, or ad-hoc queries, a relational database is usually a better fit.

Feature	DynamoDB	Traditional RDBMS
Query model	Key-value + secondary indexes	SQL with joins and aggregations
Scaling	Automatic, unlimited	Vertical (larger instance) or read replicas
Consistency	Eventual (strong optional)	ACID
Unit of billing	RCU/WCU (read/write capacity units)	Instance hour + storage

Azure Cosmos DB

Globally distributed NoSQL database with multi-model support (document, key-value, graph, column-family). Cosmos DB guarantees single-digit-millisecond latency at the 99th percentile and offers five consistency levels from strong to eventual.

Firestore (GCP)

Serverless document database with real-time synchronization, designed for mobile and web applications. It integrates with Firebase for authentication, hosting, and serverless functions.

Caching

Caching layers sit in front of databases to reduce latency and offload read traffic.

Service	Platform	Engines	Use Case
ElastiCache	AWS	Redis, Memcached	Session store, query cache, rate limiting
Azure Cache for Redis	Azure	Redis	Session store, content cache, pub/sub
Memorystore	GCP	Redis, Memcached	Session store, cache, leaderboards

Caution

Caching introduces data staleness. Decide what staleness your application can tolerate. For user sessions, milliseconds matter. For product recommendations, minutes may be acceptable. Always configure appropriate TTLs (Time To Live).

Choosing a Database

If you need…	Consider…
ACID compliance, complex joins, ad-hoc queries	RDS, Aurora, Cloud SQL, Azure SQL
Massive throughput for simple key-value lookups	DynamoDB
Global distribution with low-latency reads	DynamoDB Global Tables, Cosmos DB
Real-time sync for mobile/web clients	Firestore
Reduce read pressure on your primary database	ElastiCache, Memorystore
Flexible document model with powerful queries	MongoDB Atlas (self-managed or via provider marketplace)

Practical Example: E-Commerce Stack

User requests product page
  └── Check ElastiCache for cached product data
        ├── Cache hit → return immediately (< 1 ms)
        └── Cache miss → query DynamoDB (product catalog)
                         Query RDS (user account, order history):
                           ├── RDS Primary (writes)
                           └── RDS Read Replica (reads)

Product catalog: DynamoDB — high throughput, simple key-value access by product_id.
User accounts and orders: RDS PostgreSQL — complex queries with joins and transactions.
Session data: ElastiCache Redis — fast read/write with TTL expiration.
Search: Could be added as a separate service (Amazon OpenSearch or a dedicated search index).