A company runs a production EC2 fleet whose CPU and memory utilisation has never been measured. AWS Compute Optimizer is enabled and reports many instances as over-provisioned, with recommendations to move to smaller instance sizes and, for some, to a different instance family. The team wants to act on these data-driven findings to cut compute cost while still meeting each workload's observed demand. Which two actions directly implement the Compute Optimizer right-sizing recommendations? Select TWO.
- APurchase three-year All Upfront Reserved Instances for the current over-provisioned instance sizes to lock in the deepest committed discount immediately.
- BMove the over-provisioned production instances onto EC2 Spot Instances so the whole fleet draws from discounted spare capacity instead.
- CResize over-provisioned instances down to the smaller instance types Compute Optimizer recommends after confirming the new size meets observed CPU and memory demand. Correct
- DEnable an Auto Scaling group with a high desired capacity so additional instances absorb any demand the smaller instances cannot handle.
- EInstall the CloudWatch agent to collect memory metrics, then migrate the flagged instances to the recommended instance family informed by that fuller utilisation picture. Correct
Why A is wrong: Reserved Instances cut the rate but committing to the oversized footprint locks in waste, and buying before right-sizing is the opposite of acting on the recommendations.
Why B is wrong: Spot suits interruptible, stateless workloads and risks reclaim on a production fleet, so it changes the pricing model rather than implementing the size and family right-sizing advice.
Why C is correct: Right-sizing to the recommended smaller type removes paid-for idle capacity while still covering the measured workload, which is the core action Compute Optimizer findings drive.
Why D is wrong: Scaling out adds instances and cost rather than reducing the size of existing over-provisioned ones, so it does not implement the right-sizing recommendation.
Why E is correct: Memory is not a default EC2 metric, so the agent gives Compute Optimizer the memory data to refine family and size recommendations, and migrating to the recommended family completes the right-sizing.