A data engineer ingests daily user behavior records with Amazon Kinesis Data Streams. The stream is being throttled because some shards (hot shards) receive much more traffic than others. How should the engineer stop the throttling?

The correct solution is to use a random partition key. Kinesis Data Streams uses the partition key to route records to specific shards. If a few partition keys receive a disproportionately high number of records, those shards become "hot shards" and can lead to throttling. A random partition key ensures that records are evenly distributed across all shards, preventing hot shards and maximizing throughput. Increasing the number of shards might temporarily alleviate throttling, but if the partition key distribution remains skewed, hot shards will reappear. Limiting the number of records or decreasing record size would reduce overall data ingestion, which is not the goal; the goal is to resolve throttling while maintaining ingestion rates.