Question: Is DynamoDB a memory store?

Answer

No. DynamoDB is a durable, disk-based NoSQL database, not an in-memory store. It keeps every item on solid-state drives and replicates it across multiple Availability Zones in the Region, so data survives node failures and restarts. Typical reads and writes complete in single-digit milliseconds.

When milliseconds aren't fast enough, put a cache in front of the table:

• DynamoDB Accelerator (DAX) is an in-memory, write-through cache built for DynamoDB. It's API-compatible with DynamoDB, so you swap the client and keep your code. Cached reads return in microseconds. DAX runs on nodes that you size and pay for by the hour, and the table stays the source of truth.
• Amazon ElastiCache (Valkey, Redis OSS or Memcached) is a general-purpose in-memory store. Use it to cache computed results, sessions or data from several sources, not just DynamoDB items.
• Amazon MemoryDB is an in-memory database with a durable Multi-AZ transaction log, for workloads that want Redis-style data structures as their primary database.

How to choose:

• Need a durable primary database that scales without servers to manage? DynamoDB.
• Read-heavy DynamoDB workload with hot items, or a microsecond latency target? DynamoDB plus DAX.
• Need sorted sets, pub/sub or other in-memory data structures? ElastiCache or MemoryDB, often next to DynamoDB.

DAX only speeds up eventually consistent reads. Strongly consistent reads pass through to DynamoDB and aren't cached. A cache also adds cost and another component to operate, so add one when metrics show you need it. For a fuller comparison of a durable database with an in-memory store, see DynamoDB vs Redis.

Source: AWS DynamoDB documentation.