Affinity Statistics

[DaveB]

Would it be possible for the Project Lasso GUI to accumulate some statistics on what CPU core(s) are actually being used?

Currently we can see the set affinity, sets, %, avg, etc. but this provide no real information on what cores are actually being assigned/used by the scheduler.

In my attempts to unravel the DPC issues, I am having trouble assessing how affinities may be assigned to mitigate. The answer may ultimately lie in changes to the kernel, scheduler, or drivers, but a better understanding of why the cores are being saturated during the DPC spikes could perhaps offer some clues.

The answer may ultimately lie in changes to the kernel/drivers as there may simply be too much going on during these events.

Any ideas?

[Jeremy Collake]

Although I can't commit to anything, we'll keep this feature request in mind!

There's probably a way to collect and view historical CPU core usage with the Windows Performance Toolkit (WPR and WPA), but I don't have precise guidance for you.

[DaveB]

Although I can't commit to anything, we'll keep this feature request in mind!

There's probably a way to collect and view historical CPU core usage with the Windows Performance Toolkit (WPR and WPA), but I don't have precise guidance for you.

I am using those tools and still learning. However, WPA chokes on large volumes of data and their documentation indicates that is true. I may ultimately suggest a couple of features for WPR and/or WPA to deal with short duration events happening over long time intervals.

I liken it to attempting to find hardware timing events with an oscilloscope. Often you don't know for sure what you are looking for until hints leads you to playing games with triggering to capture. Modern scopes make it easier because they contain more memory, but still there are limits.

Thanks for considering. I understand this may not fit within the priorities of many of your users. Thank you for the time you've already spent putting Process Lasso together. It's already been helpful.

[Autumner]

I'll just add here that it'd be great for PL to provide general statistics on:

1. CPU avg (so we can know the most CPU-intensive processes)
2. Disk usage
3. RAM usage

Ideally, we'd have 1-day, 7-day and 30-day (or even custom) time-frames. Just a thought anyway, since PL already does some statistics. There are probably specific tools for this job out there already, I'm guessing.