Question

A retail company runs a SageMaker Model Monitor data quality monitoring schedule against a real-time endpoint that serves roughly 50 requests per minute. The schedule is currently configured to run hourly, and each monitoring job spins up a processing instance that runs for about 8 minutes. The ML team observes that monitoring costs have grown significantly, while their SLA only requires them to be notified of data quality drift within one business day. Which change reduces monitoring cost while still meeting the requirement?

Accepted Answer

Change the monitoring schedule to run daily instead of hourly using a cron expression.

Answer

Disable data capture on the endpoint and rely only on CloudWatch invocation metrics for drift detection.

Answer

Increase the processing instance size so each monitoring job completes faster.

Answer

Delete the baseline constraints and statistics files to skip constraint evaluation during each job.