AWS Compute Optimizer
AWS Compute Optimizer is an Amazon Web Services service that analyzes historical utilization data to identify rightsizing opportunities for Amazon EC2 instances.
Key points
- Identifies rightsizing opportunities for EC2 instances, helping reduce costs by matching instance types to actual workload needs.
- Works alongside AWS Cost Explorer, which also surfaces EC2 rightsizing recommendations.
- Also generates recommendations for other resource types, not EC2 instances alone: Amazon Elastic Block Store (EBS) volumes, AWS Lambda functions, Amazon EC2 Auto Scaling groups, Amazon ECS services on AWS Fargate, Amazon Aurora and Amazon RDS databases, and commercial software licenses.
- Distinct from Migration Evaluator, which baselines on-premises workloads and projects the cost of migrating them to AWS — Compute Optimizer only analyzes utilization data for resources already running on AWS.
- Finding classifications — resources are classified as Under-provisioned (a specification like CPU, memory, or network doesn’t meet workload demand), Over-provisioned (a specification could be sized down with no under-provisioned specification), or Optimized (all specifications meet workload demand and none are over-provisioned). These classifications apply to EC2 instances, EC2 Auto Scaling groups, EBS volumes, Lambda functions, and ECS services on Fargate.
- Requires at least 14 days of CloudWatch metric history before it can generate recommendations; recommendations refresh daily and can take up to 24 hours to first appear after a resource meets the requirements.
- Customers must opt in — Compute Optimizer is not enabled by default. Opt-in is supported for standalone accounts, member accounts of an organization, or the management account of an organization (which can then view findings across all member accounts).
- Discriminator vs AWS Trusted Advisor: Trusted Advisor runs broad best-practice checks across cost optimization, security, fault tolerance, performance, and service limits; Compute Optimizer does deep, ML-driven rightsizing analysis specifically for compute and storage resource utilization, using CloudWatch metrics history rather than static rule checks.
Pricing
- Standard Compute Optimizer recommendations (14-day lookback) are free.
- Enhanced infrastructure metrics is an optional paid feature that extends the metrics lookback period analyzed for a resource from 14 days to 93 days, giving recommendations based on longer-term trends.