Does EC2 collect metrics using a system agent? Or is it built into their virtualization?

Question

Many (most?) virtual server providers rely on metrics gathered from within the system to provide users/owners/ops with these metrics (Rackspace does this for example).

AWS does not appear to have a system-level agent, but I don't know if I'm just missing it. I think everything I see on the Basic Metrics dashboard is stuff that could be handled in the virtualization system.

So: Does CloudMetrics have a system-level agent that collects CPU, Network, or other stats that they present to users? Or is all that data collected in another layer?

Answer 1

The metrics exposed in EC2 are collected by the hypervisor. AWS, by default, has no code running on your instance. That's why memory and disk utilization are conspicuously absent from the metrics. AWS has always maintained a clean boundary between what happens inside and outside the VM.

There is an "example" agent (which is technically more than an "example" but it is referred to as such because, officially, it is unsupported even though it works fine) that can used to collect and export system level metrics that can be found at http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/mon-scripts.html. Because the CloudWatch API is open and documented, you can also write your own or even glue something together using aws-cli.

Answer 2

It's built in to their system without an agent. But there is a system manager agent and cloud watch log monitoring agent for other functionality.