Re: [compute-pressure] Data delivery and mitigation penalty algorithm issue. (#229)

> I don't think we will need to continuously trigger it, removing some of the readings will already make the mitigation work

I think we agreed on that, or we believe it s the most reasonable solution.

> Maybe just not increment
Not incrementing is still not the solution, will again trigger the penalty because change_count >= change_count_threshold.

I believe the solution is to reset change_count for the penalty source. before scheduling the penalty.
In this way, we do not interfere with the next window in case of overlap between window reset and penalty.

At the same time, the penalty can be carried again later if the count is again reaching the limit.
Other sources penalty can also be detected, since we don't invalidate the window after the first penalty on any source.



-- 
GitHub Notification of comment by arskama
Please view or discuss this issue at https://github.com/w3c/compute-pressure/issues/229#issuecomment-1693573805 using your GitHub account


-- 
Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config

Received on Friday, 25 August 2023 15:52:32 UTC