Question: Is DynamoDB OLTP or OLAP?

Answer

OLTP. DynamoDB is an online transaction processing database: it handles large numbers of small, fast reads and writes of individual items, or of groups of items that share a partition key, with consistent single-digit-millisecond latency at any scale. Shopping carts, user profiles, sessions, game state and order records are typical workloads.

It isn't an OLAP engine. There are no server-side joins or SQL GROUP BY aggregations, and analyzing a whole table means a Scan that reads every item and consumes read capacity as it goes.

For analytics, copy the data to a system built for it, without loading the production table:

• Export to S3. Full and incremental exports require point-in-time recovery and use no read capacity. Query the files with Amazon Athena, AWS Glue or Spark.
• Zero-ETL integrations. Replicate a table into Amazon Redshift, Amazon SageMaker Lakehouse or Amazon S3 Tables (through AWS Glue), or into Amazon OpenSearch Service for search. AWS states that zero-ETL integrations have no impact on production workloads running on DynamoDB.
• Streams. Process DynamoDB Streams with Lambda to maintain running totals or feed a near-real-time pipeline.

A common architecture keeps DynamoDB as the OLTP system of record and sends a copy to a warehouse or data lake for reporting. DynamoDB vs Redshift compares the two roles, and export DynamoDB data to S3 shows the export options.

Source: AWS DynamoDB documentation.