This might be a total noob question, but that's what I am, so here we go: I'm trying to understand how tasks suspension works in BOINC with WCGrid. There are options as to the maximum number of cores used, and an option to suspend if a certain number of non-BOINC processes are run. To me this seems to suggest that BOINC will either run or suspend ALL tasks, whereas intuitively I would think that on a multiprocessor system suspending only a subset of running tasks would be the way to go, e.g. run on a maximum of 100 out of 120 cores, and if 21 cores are used for other stuff, suspend one task, if 30 are used, suspend 10 etc.. Is my understanding correct?

This might be a total noob question, but that's what I am, so here we go: I'm trying to understand how tasks suspension works in BOINC with WCGrid. There are options as to the maximum number of cores used, and an option to suspend if a certain number of non-BOINC processes are run. To me this seems to suggest that BOINC will either run or suspend ALL tasks, whereas intuitively I would think that on a multiprocessor system suspending only a subset of running tasks would be the way to go, e.g. run on a maximum of 100 out of 120 cores, and if 21 cores are used for other stuff, suspend one task, if 30 are used, suspend 10 etc.. Is my understanding correct?

There is also the option of running x% of CPUs hence 100 out of 120 cores but this is a single limit, not banded.

The main thing to note, however, is that Boinc runs at the lowest possible priority and will automatically stop if any normal process starts.

Use of the suspend function is all or nothing as I understand it. There is I believe a request at git-hub where the code is maintained and developed to suspend some cores depending on cpu usage but that request is not being actively worked on.

In theory BOINC even without the suspend when ... features ticked should give up time gracefully to other work being done on the computer. However, in practice, certainly on old machines such as the core2 duo I recently replaced running BOINC using both cores would sometimes make it unresponsive for minutes at a time. On my Ryzen, even running all 16 threads using BOINC I have not seen this.

So I'm guessing low priority means niceness 19? Our institute has a policy of requesting that all users run their computations at niceness 19 to ensure smooth SSH and RStudio sessions. I guess that would be a problem then with BOINC?

So I'm guessing low priority means niceness 19? Our institute has a policy of requesting that all users run their computations at niceness 19 to ensure smooth SSH and RStudio sessions. I guess that would be a problem then with BOINC?

The BOINC client and manager run at normal priority or nice level 0. The science applications from projects run at whatever level the developers of those applications choose. In general cpu applications run at nice level 19 and gpu applications run at nice level 10.