Re: [compute-pressure] Add an implementation-defined change threshold exceeded check (#180)

@kenchris thanks for the comments. I pushed an update, PTAL.

This was a bit hard to define in abstract without defining the concrete algorithm normatively, which I don't want to do. Thus I proposed that short two-step algorithm and an accompanying note that tries to explain what's the problem and what implementers should do to improve. I think the significance of the changes to contributing factors need to be considered collectively. That is, a significant change in one metric may not exceed the threshold for the entire system if other metrics remain within their own bounds.

Also, I wanted this to remain implementation-defined and thus I tried to avoid saying explicitly "if CPU utilization fluctuates around a boundary" to not steer implementers to think that is the metric to use.

Please take another look. Further suggestions welcome.

GitHub Notification of comment by anssiko
Please view or discuss this issue at using your GitHub account

Sent via github-notify-ml as configured in

Received on Wednesday, 1 February 2023 16:35:28 UTC