Question

A media company is building a new real-time clickstream analytics platform. During major live events, ingestion spikes from 2,000 to over 400,000 events per second within minutes, then drops back down. The analytics team requires that events from the same user session be processed in strict order, and the platform must retain raw events for 7 days so they can be reprocessed if a downstream bug is discovered. The solution must scale automatically without manual capacity planning and minimize operational overhead. Which approach best meets these requirements?

Accepted Answer

Ingest events into Amazon Kinesis Data Streams using the on-demand capacity mode, partition records by session ID, and configure a 7-day data retention period.

Answer

Publish events to an Amazon SQS standard queue, use session ID as the message deduplication ID, and process with Lambda; store raw events in S3 for 7 days.

Answer

Send events to Amazon Kinesis Data Firehose delivering directly to S3, and run Athena queries against the last 7 days of stored events for reprocessing.

Answer

Ingest events into Amazon Kinesis Data Streams using provisioned mode with 500 shards permanently allocated, partition by session ID, and set a 7-day retention.