Flash card
AWS Compute Optimizer: right-sizing recommendations derived from CloudWatch utilisation history.
- Analyses observed CloudWatch metrics and recommends a specific instance type, volume type, or memory setting, rather than pointing at a category of waste.
- Covers EC2 instances, Auto Scaling groups, EBS volumes, Lambda functions, and ECS services on Fargate.
- Needs a history to work from, so a workload that has just launched or just changed shape produces weak recommendations.
- Opt-in per account, and enrollable for a whole organisation from the management account.
- Memory is not an EC2 metric by default, so recommendations sharpen considerably once the CloudWatch agent is reporting memory utilisation.
Pick it when
Pick it over Trusted Advisor when the question is “what size should this be”. Compute Optimizer answers with a specific target based on measured utilisation; Trusted Advisor’s cost checks flag idle and underused resources without sizing the replacement.
It's the wrong answer when
It is the wrong answer when nothing has been measured yet, when the resource type is outside its coverage, or when the scenario wants commitment-based savings. Recommending a smaller instance does nothing about a workload whose spend problem is on-demand pricing for steady load, which is a Savings Plans question.