I have a pipeline that runs some computationally intensive tasks on a Linux machine. The script that launches these checks the current load average and waits if it is above a certain threshold. I would like to ask for help determining what that threshold should be.
I am thinking of asking on SF because i) this is running on a professional environment, a ~20 CPU server (I say approximately because this is a VM and the number of CPUs varies, which is why I need to vary my threshold accordingly) and ii) this is more of a sysadmin issue than a programming one; I know how to set the average, I need help in choosing what the threshold should be. We currently allow it to be well above the number of cores and it doesn't seem to be causing any trouble.
I am not, however, a professional sysadmin as such. On the other hand, I work in a small startup and I am as much of one as anyone else around. Since I am aware of SF's scope, I though I'd ask before actually asking.
So, would such a question from a non-professional sysadmin who is nevertheless performing profesional sysadmin tasks at the moment be on topic here?
$(grep -c processor /proc/cpuinfo) x NwhereN>=1and would like some help in figuring out what valueNshould take. My limited understanding of how load works suggests that N should always be 1 but my equally limited experience shows that >1 can also work. Hence the question. Is there no commonly used heuristic for this?