Skip to content

PerfMonComps: Restructuring/Improving various aspects of PerfMonMTSvc

This MR restructures/(hopefully 😄)improves various aspects of PerfMonMTSvc. To name a few:

  • Added necessary flags to athena so that users can easily use the service
  • Improved the service configuration, implemented new flags and bells and whistles
  • Unified the parallel/serial component monitoring code (see below)
  • Added job configuration step as a new snapshot, renamed Event Loop to Execute
  • Cleaned up a number of methods/members that are not necessary, implemented initializations etc.
  • Improved the results reporting in the log

The most major change is how we collect component-level data. Now, we collect and store data from each slot separately. Once the job is done, before reporting, results are aggregated over all slots before reporting.

There are still a number of things that can/should be improved. These will go in another MR. Since the service is not yet used in production code, these are not super urgent can be addressed as we go along. I wanted to submit this to make review easier.

cc: @haozturk

Merge request reports