Multiple EC2 instances in a dev environment run a resource-heavy app. The SysOps administrator needs an automated way to stop these instances when they are idle. Which method will meet this need?

The most effective method is to create a CloudWatch alarm that stops EC2 instances when average CPU utilization stays below a threshold (e.g., 5%) for a sustained period (e.g., 30 minutes). This directly measures instance activity and allows for automated actions. Inspecting CloudTrail logs for EC2 API activity is not suitable because it monitors API calls to EC2, not the activity within the instances themselves. An idle instance might not generate EC2 API calls, but this doesn't reflect its workload. Using VolumeReadBytes is an indirect and less reliable indicator of idleness. An instance could be performing CPU-intensive tasks without significant disk I/O, or it could be genuinely idle but still have some background I/O. AWS Config is primarily for evaluating resource compliance and changes, not for monitoring real-time operational metrics like CPU utilization to determine idleness for automated stopping.