In an autoscaling environment, you just need to keep your metric paths consistent. Then all you need is wildcard queries. E.g., CPU userspace cycles on all your instances should be like:

    provider.region.vpc.group.instance-id.os.cpu.user

E.g.:

    aws.us-west-1.prod.web-front.i-4df567a9.os.cpu.user

Then wildcard queries will provide you aggregate stats at any level you like.

It's really a feature of how Graphite queries work.

--------------------

All that being said, I agree that Grafana is pretty awesome.