AWS Certified Data Engineer - Associate

AWS Certified Data Engineer - Associate · free preview Question 1 of 10

A fleet of 80,000 IoT devices sends telemetry to an Amazon Kinesis Data Stream. The producer currently uses deviceType as the partition key. During peak traffic, CloudWatch shows WriteProvisionedThroughputExceeded errors on a small number of shards, while most shards have low utilization. The application must preserve ordering per device, but it does not require ordering across devices.

Which change best addresses the ingestion bottleneck while preserving the required ordering?

Batch records into larger PutRecords requests while continuing to use deviceType as the partition key.

Use deviceId as the partition key and increase stream capacity, or use on-demand capacity mode if traffic is variable.

Use a random UUID as the partition key for every record.

Continue using deviceType as the partition key and enable enhanced fan-out for all consumers.