DynamoDB is Amazon's fully managed key-value and document database. You don't pick servers, install anything, run repairs or plan failovers. You create a table, choose a key, and it serves single-digit-millisecond reads at almost any scale. Amazon's own retail systems lean on it during Prime Day: AWS reported a peak of 192 million requests a second for Prime Day 2026.
The price of that simplicity is that you must know your access patterns up front, and model your data around a key. If you've read the Cassandra lesson, much of this will feel familiar: DynamoDB descends from the same 2007 Dynamo paper, though the modern service works differently inside. Module 14's partitioning ideas are the foundation here.
What it's best at
- Known, key-based access at any scale: "get order o_9912", "latest 20 orders for restaurant r_42".
- Spiky or unpredictable traffic: on-demand mode scales up and down with no capacity planning.
- Teams that don't want to run databases: no patching, no replicas to babysit, backups and point-in-time restore built in.
- Serverless designs: it pairs naturally with Lambda and pay-per-request pricing.