Question: How is data stored in DynamoDB?

Answer

DynamoDB stores data as items in tables, with each item containing named attributes. Underneath, AWS stores the data on SSD-backed partitions and automatically replicates it across multiple Availability Zones within the table's Region.

Every item has a primary key: either a partition key alone or a partition key plus a sort key. DynamoDB hashes the partition-key value to locate the data. Items sharing that value form an item collection and are ordered by sort key. AWS manages partition allocation and splitting; you do not select the servers.

Attributes can contain strings, numbers, binary data, booleans, nulls, maps, lists or sets. Items in the same table can have different non-key attributes, so a document does not need to be flattened. The complete item, including attribute names, must fit within 400 KB.

GetItem retrieves an item by its full key. Query reads a key range from a table or index; Scan examines items across the table. Filters discard already-read items and do not substitute for an index.

A normal table is not automatically copied to other Regions. Enable global tables when you need regional replicas. See partition-key design and supported data types for the logical model.

Source: AWS partitions.