DynamoDB designed around your access patterns

When access patterns are known, volumes are high and you want minimal operations. If you expect frequent new queries or complex reporting, a relational database is usually easier.

By choosing the right capacity mode, designing keys to avoid scans, setting expiry on temporary data and monitoring usage so surprises are caught early.

It stores several entity types in one table, using carefully chosen partition and sort keys so related items can be fetched in a single query. It is efficient but needs the access patterns agreed before building.

They let you query a table by attributes other than its primary key, each with its own key structure. We add them for specific access patterns, bearing in mind each one adds write cost and is eventually consistent.

We enable point-in-time recovery, which allows restoring a table to any moment within the retention window, and add on-demand backups before risky changes. Restores create a new table, so we plan the cut-over.

Yes. It offers transactional reads and writes across multiple items and tables, with all-or-nothing behaviour. They cost more than normal operations, so we reserve them for cases that need them.

Hot partitions happen when too many requests hit the same partition key, such as a single popular customer or today's date. We spread load with better key design, for example by adding a suffix to high-traffic keys.

Streams capture every change to a table, which can trigger Lambda functions for notifications, search indexing, audit logs or syncing to analytics. This keeps side effects out of the main request path.

Not efficiently inside DynamoDB itself. We export data to S3 and query it with tools such as Athena, or stream changes into a warehouse, so reporting never scans the live table.

Yes, but it is a redesign rather than a copy, because tables must be restructured around access patterns. We list every query the application needs before designing keys and moving the data.