Finding Inefficiently Used ML Compute and EBS Resources with AWS Compute Optimizer Recommendations

Why the Answer Is Correct

The requirement has two parts, identifying inefficiently used resources and generating recommendations to reduce their cost, and it must be done with the least development effort. AWS Compute Optimizer is a managed service that analyzes the actual utilization metrics of EC2 instances, EBS volumes, and Lambda functions and then produces rightsizing and cost optimization recommendations automatically, satisfying both halves without the engineer writing any analysis code. The vote was unanimous at 100 for D. Saransundar described Compute Optimizer as finding wasted resources in EC2 and EBS and suggesting easy ways to save money and boost performance, and aws_Tspy pointed to the least-development-effort framing as the decisive clue.

Why the Other Options Are Wrong

Creating code to evaluate each instance's memory and compute usage (A) is the high-effort path the question explicitly rules out, since it requires building, running, and maintaining custom monitoring and analysis logic across EC2, Lambda, and ECS. Adding cost allocation tags and activating them in AWS Billing and Cost Management (B) answers a different question, namely who is spending what, and while tags are useful for chargeback they do not detect over-provisioned or idle resources and generate no rightsizing recommendations. Checking CloudTrail event history for resource creation (C) only reveals when resources were created and by whom, not how heavily they are used, so it cannot identify waste or produce cost recommendations.

Community Comment Notes

The community was unanimous at 100 for D. Saransundar identified the two-part fit precisely, noting that Compute Optimizer finds wasted resources in EC2 and EBS and suggests concrete ways to save money. aws_Tpozit identified the keyword as identifying inefficient resources while generating recommendations with minimal effort, and then explicitly dismissed option A on the grounds that writing custom evaluation code is the opposite of minimal effort. The reasoning shows the question is testing the difference between a managed recommendation service and manual or attribution-only approaches.