Reduce Kinesis Iterator Age by Increasing Shards

Quick Answer

Iterator age is Kinesis's way of surfacing how far behind a consumer has fallen: it measures the gap between the timestamp of the oldest unread record in a shard and the current time, so a rising iterator age is a direct signal that records are arriving faster than the consumer can read them, not that anything is wrong with the records themselves. Because each shard in a Kinesis Data Stream has a fixed read throughput ceiling, the total amount of data a consumer fleet can pull from the stream per second is capped by how many shards exist, no matter how efficiently the consuming application code is written. Increasing the number of shards raises that ceiling directly, giving the custom application more parallel partitions to read from simultaneously, which increases aggregate throughput and lets the consumer catch up to the incoming IoT data rate, bringing iterator age back down. This is a different lever from optimizing the consumer application's own processing logic, even a perfectly efficient consumer is still bounded by the total read capacity the shard count provides, so when the bottleneck is genuinely throughput rather than processing inefficiency, adding shards is the fix that scales with the actual constraint. Whenever a Kinesis-based scenario describes a consumer falling behind with a rising iterator age, and doesn't point to a specific application-level inefficiency, treat shard count as the first thing to check, since it directly controls the stream's total read capacity.