Error: DynamoDB query is slow

A slow Query is not necessarily scanning the whole table. Query uses the selected table or index key, but a large item collection, selective post-read filter, throttling retries, or client overhead can make the complete operation slow. Measure one service call and the full pagination loop separately.

1. Record client elapsed time, SDK attempts, response size, Count, ScannedCount, and ConsumedCapacity. Compare client latency with SuccessfulRequestLatency; SDK retries and network delays add time outside a successful service request.
2. Move useful sort-key restrictions into KeyConditionExpression. A FilterExpression removes items after reading them, so low Count relative to ScannedCount can reveal an inefficient access pattern.
3. Check ThrottlingReasons and the resource ARN. Hot partitions and GSI bottlenecks need targeted fixes, not an automatic table-wide capacity increase.
4. Reuse SDK clients and connections, use the intended Region, and avoid fetching every page when the application needs only the first page.
5. Design a GSI for a different access pattern when justified. Caching can help repeated reads if the application accepts its invalidation and consistency behavior.

ProjectionExpression can reduce response payload and client processing, but does not reduce read capacity for the same evaluated table items. Treat each proposed change as a measured fix rather than assuming table size alone is the cause.

See Query access patterns and AWS documentation.