Other Databases — Concept (Neptune, DocumentDB, Keyspaces, QLDB, Timestream, MemoryDB)

These are AWS's purpose-built databases. SAA-C03 rarely deep-dives but tests picking the right one.

Neptune

• Managed graph database.
• Two query languages: Gremlin (property graph) + SPARQL (RDF).
• Use cases: social networks, recommendation engines, fraud detection, knowledge graphs.
• Multi-AZ, up to 15 read replicas.
• Neptune Streams for change capture.

DocumentDB

• Managed MongoDB-compatible document DB (JSON).
• Drop-in for many MongoDB apps.
• Multi-AZ, scales reads with replicas.
• Up to 15 replicas, storage auto-scales.
• Best when you need MongoDB API without running it.

Keyspaces (for Apache Cassandra)

• Managed Cassandra-compatible, wide-column NoSQL.
• Serverless capacity (on-demand) or provisioned.
• CQL (Cassandra Query Language).
• Use for time-series, IoT, large-scale logs (legacy Cassandra workloads).

QLDB (Quantum Ledger Database)

• Immutable, cryptographically verifiable ledger.
• Centralized (single trusted owner) — not blockchain.
• Use for system-of-record audits, supply chain, financial records.
• Append-only, every change cryptographically chained.

Note: AWS announced QLDB end-of-life (scheduled). Still appears in exam questions today; verify on exam day. For new builds AWS suggests Aurora PostgreSQL.

Timestream

• Managed time-series database.
• Stores recent data in-memory ("memory store"), older in cheaper "magnetic store".
• SQL-like queries with time-series functions.
• Use for IoT, DevOps metrics, industrial telemetry.

MemoryDB for Redis

• Managed Redis-compatible in-memory durable database (Multi-AZ transaction log).
• ElastiCache Redis is a cache; MemoryDB is a primary database with the same API.
• Use when you need ultra-low-latency reads AND persistence as a real DB.

Picking the right one

Common exam scenarios

1. "Recommendation engine traverses user-product graph" → Neptune.
2. "Lift-and-shift a MongoDB workload" → DocumentDB.
3. "Existing Cassandra app, want managed" → Keyspaces.
4. "Immutable, cryptographically verifiable order history" → QLDB (today; verify exam currency).
5. "Millions of IoT sensor readings, query last 30 days" → Timestream.
6. "Redis API but as primary durable DB" → MemoryDB.

Exam tip

Pick by workload model, not by perceived performance:

• relational? → RDS / Aurora
• key-value, massive scale? → DynamoDB
• graph? → Neptune
• doc (MongoDB)? → DocumentDB
• wide-column (Cassandra)? → Keyspaces
• time-series? → Timestream
• ledger? → QLDB (legacy)
• in-memory cache? → ElastiCache
• in-memory DB (durable)? → MemoryDB.

References

• https://aws.amazon.com/products/databases/