Metrics identification and collection

[knarayan]

Identify the specific metrics to be used for the node CPU utilization vs. power modeling, and find ways to collect those metrics.

[husky-parul]

Nov 6

Currently we are getting constant metric values from Kepler.

• Use the latest Kepler image to see if utilization and power changes
• Modify workloads

Metrics we are collecting right now

1. Node:
   For CPU utilization node_cpu_seconds_total (from kube metrics)
   For Power kepler_node_package_joules_total (from kepler)

For Kepler we were seeing power metrics from only one node. Also we are only getting RAPL power. We do not see other power metrics from Kepler.