MLA-C01 Practice Question 99: Idle Endpoint Recommendations
AWS Certified Machine Learning Engineer – Associate (MLA-C01) · Test 10 · question 9 of 10
A FinOps team suspects that data science teams across 12 AWS accounts have left SageMaker AI real-time endpoints running after experiments ended. The team wants a regularly refreshed list of endpoints that have received no traffic for the past two weeks, with an estimate of the monthly savings from removing them, without writing scripts or building dashboards in each account.
What should the FinOps team use?
Show answer and explanation
Answer:
AWS Compute Optimizer produces idle recommendations for SageMaker AI endpoints. An endpoint is flagged as idle when it had zero invocations over the 14-day default lookback period. Recommendations refresh daily, include estimated monthly savings, and suggest deleting endpoints that are no longer needed or using scale-to-zero for inference components. That is the list the FinOps team wants, with no custom code (option B).
- A is wrong because Cost Explorer shows how much hosting *costs*, not whether an endpoint received any traffic. An expensive endpoint can be busy, and an idle one can look ordinary. - C is wrong because Inference Recommender benchmarks a model to choose an instance type for expected traffic. It does not detect endpoints nobody is calling, and it would have to be run model by model. - D is wrong because building and reviewing per-endpoint dashboards in 12 accounts is the manual work the team wants to avoid, and dashboards give no savings estimate.
Reference: Viewing idle resource recommendations - AWS Compute Optimizer