The correct answer is: Use Amazon DynamoDB with DAX to serve frequently accessed data. Export the table contents to Amazon S3 using built-in export functionality, and run analytics queries using Amazon Athena.

This solution satisfies both latency and analytics requirements with minimal operational overhead. DynamoDB is optimized for key-value workloads and delivers fast performance, while DynamoDB Accelerator (DAX) further reduces read latency to microseconds by caching results in memory. For historical analysis, DynamoDB offers native table export to Amazon S3 in a serverless and maintenance-free way. Once in S3, the data can be queried on demand using Amazon Athena, which provides a cost-effective solution for ad hoc querying without provisioning infrastructure.

The choice to use Amazon RDS and manually export data to S3 introduces unnecessary operational complexity. RDS can support transactional workloads, but it lacks microsecond read latency and requires a custom script or ETL job to transfer data, adding maintenance overhead.

The option to store all data in Amazon S3 and transition it to Glacier Deep Archive fails to meet the real-time latency requirements of the application. S3 combined with Athena is effective for large-scale analytics, but it does not support sub-millisecond access patterns necessary for real-time leaderboard updates.

The choice to use DynamoDB with Kinesis Data Streams and Firehose to stream data into S3 adds architectural complexity and is overengineered for the stated need. This pattern suits continuous ingestion pipelines, but it requires multiple managed services and does not simplify the one-time export or querying process as effectively as native table export and Athena.

This question is valuable because it tests your ability to architect hybrid solutions that balance operational speed with long-term data analytics. Many candidates struggle to combine real-time and batch-processing use cases into a unified, low-maintenance design.

To answer questions like this, pay attention to keywords such as sub-millisecond latency, one-time queries, and least operational overhead. When you see those terms, think of serverless, managed services like DAX, S3, and Athena, and avoid answers that include manual scripting or overly complex data pipelines unless there's a specific requirement for streaming or real-time ingestion.

Full AWS Practitioner Certification Question