Why it matters
Right-sizing is one of the fastest ways to reduce AWS costs, but doing it manually across hundreds of workloads is hard.
AWS Compute Optimizer uses machine learning and utilization metrics to recommend better instance types, sizes, and configurations so that resources match actual usage instead of guesses.
What Compute Optimizer analyzes
Compute Optimizer can analyze:
- EC2 instances and Auto Scaling groups – CPU, memory (via CloudWatch), network, and disk utilization
- EBS volumes – IOPS and throughput usage
- Lambda functions – memory and duration patterns
- ECS services on Fargate – CPU and memory reservations vs. actual use
This provides a centralized view of over-provisioned and under-provisioned workloads.
How to enable it
- Open the AWS Compute Optimizer console.
- Choose Account opt-in (or Organization opt-in if using AWS Organizations) to analyze supported resources.
- Confirm that permission is granted for Compute Optimizer to read CloudWatch metrics and configuration data.
- Wait for the initial analysis (typically a few hours, based on available historical metrics).
Once enabled, recommendations are refreshed regularly without additional configuration.
How to use the recommendations
- Filter by service, account, or tag to focus on specific teams or environments
- Start with over-provisioned resources to capture safe, immediate savings
- Review the confidence rating and expected savings before making changes
- Test changes in non-production first for critical workloads
Compute Optimizer is especially powerful when combined with:
- The Basics in this section (right-sizing, Auto Scaling, turning off idle environments)
- Budgets and Cost Anomaly Detection to monitor spend and catch regressions
Best practices
- Enable Compute Optimizer at the organization level so all member accounts are included
- Make it part of a regular review cycle (for example, monthly cost reviews)
- Document a simple workflow: review → validate → implement recommendations