Amazon DynamoDB
Serverless, distributed NoSQL key-value and document database delivering single-digit millisecond latency at any scale.
Serverless, distributed NoSQL key-value and document database delivering single-digit millisecond latency at any scale.
Coming from Google Cloud?
Both offer high-throughput NoSQL document/key-value storage; Firestore matches DynamoDB's serverless document model, while Bigtable matches its ultra-high-throughput partition sharding.
- Fully managed NoSQL storage with automatic scaling and high availability.
- Single-digit millisecond latency and point-in-time recovery.
- Change data capture streams for event-driven serverless triggers.
- Query Model: DynamoDB strictly enforces query patterns via Partition and Sort keys (no multi-attribute indexes without GSIs); Firestore allows ad-hoc composite indexes.
- Capacity Model: DynamoDB offers explicit RCU/WCU provisioned mode alongside On-Demand; Firestore is strictly on-demand serverless.
- Global Replication: DynamoDB Global Tables provide multi-region active-active writes; Firestore offers multi-region active-passive configurations.
What is Amazon DynamoDB?
Amazon DynamoDB is a fully managed, serverless, key-value and document database designed to run high-performance applications at any scale. DynamoDB offers built-in security, continuous backups, automated multi-region replication, in-memory caching with DAX, and automated data export to S3. It scales horizontally to millions of requests per second with single-digit millisecond latency.
Decision Guide: When to Choose This Service
Use Amazon DynamoDB for high-throughput operational workloads requiring predictable single-digit millisecond latency, shopping carts, gaming leaderboards, session stores, mobile app backends, and event-driven serverless architectures.
Do NOT use DynamoDB for complex relational queries requiring multi-table SQL joins, ad-hoc aggregations, or analytical full-table scans (use Amazon Aurora or Redshift).
Core Architectural Pillars
Composite primary key enabling $O(1)$ partition lookups and sorted range queries.
Instant auto-scaling pay-per-request vs. planned RCU/WCU provisioning.
Asynchronous secondary index with independent partition and sort keys.
Time-ordered 24-hour change data capture (CDC) log powering event-driven Lambda triggers.
Multi-region active-active replication with sub-second cross-continent propagation.