Question

A developer is building a real-time analytics platform that reads from a single Amazon Kinesis Data Stream. Currently three separate Lambda consumer applications read from the same stream using standard (shared-throughput) consumers, and the team is seeing increased read latency and GetRecords throttling as more consumers are added. The team wants each consumer to receive records with sub-200ms latency and its own dedicated read throughput without provisioning additional shards. What should the developer do?

Accepted Answer

Register each consumer application as an enhanced fan-out consumer so each gets a dedicated 2 MB/sec per-shard read throughput pipe using the SubscribeToShard API.

Answer

Increase the number of shards in the stream so that each consumer has its own dedicated set of shards to poll.

Answer

Switch the consumers to use the GetRecords API with a larger batch size and a shorter polling interval to reduce latency.

Answer

Migrate the stream to on-demand capacity mode so Kinesis automatically scales read throughput per consumer.