Exam relevance: Domain 3 (25%). Compute Optimizer uses ML to recommend optimal resource configurations. Know supported resources, recommendation types, and how it differs from Cost Explorer rightsizing.
1. Overview
AWS Compute Optimizer analyzes historical utilization metrics using machine learning and recommends optimal AWS resource configurations to reduce cost and improve performance.
Compute Optimizer vs Cost Explorer Rightsizing:
| Feature | Compute Optimizer | Cost Explorer Rightsizing |
|---|---|---|
| ML-based | Yes | No (threshold-based) |
| Supported resources | EC2, ASG, Lambda, ECS Fargate, EBS, RDS | EC2 only |
| Performance risk | Shows risk of under-provisioning | Not shown |
| Enhanced metrics | Yes (CloudWatch Agent memory) | Yes (CloudWatch Agent) |
| Cost | Free (standard), paid (enhanced) | Free |
2. Supported Resources
| Resource | What It Optimizes |
|---|---|
| EC2 instances | Instance type, size |
| EC2 Auto Scaling Groups | Instance type, size |
| EBS volumes | Volume type, size, IOPS |
| Lambda functions | Memory size |
| ECS services on Fargate | CPU and memory |
| RDS DB instances | Instance class (preview) |
3. Recommendation Types
Each recommendation includes:
- Current configuration and its cost
- Recommended configuration and estimated cost
- Performance risk: risk of the recommendation being under-provisioned
- Savings opportunity: estimated monthly savings
- Finding: Over-provisioned, Under-provisioned, Optimized, Not enough data
4. Enhanced Infrastructure Metrics
- Paid feature: uses 3 months of metrics instead of 14 days
- More accurate recommendations for workloads with seasonal patterns
- Requires opt-in per account or organization-wide
5. Organization-Wide Recommendations
- Enable Compute Optimizer at the organization level from the management account
- View recommendations across all member accounts in one place
- Aggregate savings opportunities across the organization
6. Exam Scenarios
Scenario 1: A company wants ML-based recommendations to rightsize EC2 instances, Lambda functions, and EBS volumes across all accounts. What is the solution?
Enable AWS Compute Optimizer at the organization level. It analyzes utilization metrics for EC2, Lambda, EBS, ECS Fargate, and ASGs using ML. View consolidated recommendations across all accounts from the management account.
Scenario 2: Compute Optimizer recommends downsizing an EC2 instance but the team is concerned about performance impact. What information does Compute Optimizer provide?
Compute Optimizer includes a performance risk indicator for each recommendation — Low, Medium, or High. High risk means the recommended instance may be under-provisioned for peak workloads. The team can choose a recommendation with lower performance risk at a slightly higher cost.
7. Exam Tips
- Compute Optimizer = ML-based, covers EC2, Lambda, EBS, ECS Fargate, ASG
- Cost Explorer rightsizing = EC2 only, threshold-based
- Performance risk is unique to Compute Optimizer — shows risk of under-provisioning
- Enable at organization level for consolidated recommendations
- Requires opt-in — not enabled by default
- Enhanced metrics (paid) = 3 months of data for seasonal workloads