A company wants to lower costs for containerized ML workloads running on EC2 instances, Lambda, and an ECS cluster. EC2 and ECS use EBS volumes to store predictions and artifacts. The ML engineer must identify underutilized resources and get recommendations to reduce cost with minimal development effort. Which option should they choose?

AWS Compute Optimizer analyzes the configuration and utilization metrics of your AWS resources, including EC2 instances, Lambda functions, and EBS volumes, to recommend optimal resource configurations. This directly addresses the need to identify underutilized resources and provides cost-saving recommendations with minimal effort, as it's an automated service. Writing custom code to analyze memory and CPU utilization is time-consuming and requires significant development effort, which goes against the requirement for minimal effort. Applying cost-allocation tags helps track costs but doesn't identify underutilized resources or provide optimization recommendations. Inspecting CloudTrail event history shows when resources were created but offers no insight into their current utilization or potential for cost reduction.