TOPIC CLOUD
DynamoDB serverless NoSQL database with single-digit millisecond latency
IN 10 SECONDS
DynamoDB is a fully managed key-value and document database. You define a table with a primary key (partition key + optional sort key), and DynamoDB automatically distributes data across partitions. It scales horizontally — throughput is provisioned or on-demand. It's the backbone of countless serverless architectures at Amazon.
GOTCHA A hot partition (heavily accessed key) can throttle your entire table. Design your partition keys for high cardinality — use a `userId` or a composite key, not a `status` field with only 3 values. DynamoDB limits a single partition to 3000 RCU / 1000 WCU.
HOW DYNAMODB SERVES A QUERY
01 Application calls `GetItem` with a primary key: `{ 'userId': 'user_abc123' }`.
02 DynamoDB hashes the partition key to determine which partition (physical storage node) holds the data.
03 Partition retrieves the item from SSD storage within the partition and returns it.
04 Consistency by default, reads are eventually consistent (1s replication lag). Strongly consistent reads cost more but return the latest data.
POKE IT YOURSELF
aws dynamodb describe-table --table-name my-table — inspect table metadata, indexes, and capacity
aws dynamodb scan --table-name my-table — scan all items (expensive on large tables)
Drill this topic →
~85 sec read