WIP Adding a baseline kube-burner job

[jtaleric]

Run kube-burner to capture baseline of usage on the host.

TODO : Today, the evaluation uses namespace=kepler however if we want to baseline, we should just capture total node cpu usage for before and after. We can also capture the kepler namespace just to understand the overhead.

[jtaleric]

@dave-tucker I am not in love with how I am storing the results -- just generating the json, but doing nothing with it.

How do we want to compare to the baseline? Just store the artifact or actually build a diff?

[dave-tucker]

@jtaleric I'll take raw data for now. We have a task to work on better reporting.
Adding a new field to the stress test markdown report looks like it will be a huge pain in the backside 😓
See: ./util/calc_mean_sd.awk for how new entries get appended to that table each run.

I'd like to see something like:

mean cpu no kepler	mean cpu with kepler	mean cpu consumed by kepler pods	overhead
50%	75%	1%	24%

where overhead = mean cpu with kepler - mean cpu consumed by kepler - mean cpu no kepler

[jtaleric]

@jtaleric I'll take raw data for now. We have a task to work on better reporting. Adding a new field to the stress test markdown report looks like it will be a huge pain in the backside 😓 See: ./util/calc_mean_sd.awk for how new entries get appended to that table each run.

I'd like to see something like:

mean cpu no kepler mean cpu with kepler mean cpu consumed by kepler pods overhead
50% 75% 1% 24%
where overhead = mean cpu with kepler - mean cpu consumed by kepler - mean cpu no kepler

ack - I will go ahead and make this test run longer too. We can pick up a separate task to report.