An application stores data in DynamoDB. Some queries are slow because they filter on an attribute that is neither the table partition key nor sort key. The dataset will grow significantly. What change will improve query performance for that attribute?

Creating a Global Secondary Index (GSI) with the queried attribute as its partition key is the most effective solution. GSIs allow you to query non-key attributes efficiently, as DynamoDB builds a separate index that can be queried independently. This avoids full table scans, which are inefficient and slow, especially with growing datasets. Increasing the request page size or using parallel scans are still performing scans, which are not ideal for performance-critical queries on non-key attributes. Enabling read capacity auto scaling only addresses throughput issues, not the underlying inefficiency of the query itself.