Background

Currently kepler has a high overhead as pointed out in the perf test

User Story

1. As an OpenShift customer using Power Monitoring on Bare metal or VM, it should not consume more than the following resources

• CPU : 1%
• Memory :
• Kube Api :
• UWM / Incluster Prometheus:
• Average core usage per Node:

1. As an Openshift customer, I should be able to control / filter the amount of metrics that kepler produces to reduce the overhead on the monitoring stack

• e.g. I only want to know kepler_process_package_joules_total and not core, uncore dram etc ..