MetricStore should
- downsample 15 minute granularity data to 1 hour buckets
- upsample 15 minute granularity data to 5 minute buckets with interpolated data
- aggregate vm data to a host
- graph the max of cpu usage of all vms
- graph the number of vms running on a host