I'm running due diligence over BOINC and was wondering how the client handled resource control on a Linux machine. Are the limits absolute or are resources dynamically controlled? Could I set a rule whereby a project consumers 80% of resources on a machine, but that drops down to 20% when other processes on the machine require it?

If anyone could point me to that part in the source code that does the resource control, that would be appreciated.

That fine-featured control does not exist in the client. The closest you can approximate is setting the percentage of the cpu used globally by the client and then setting some exclusions parameters to reduce the cpu availability when specific applications running on the host are detected which would knock down the current cpu usage by the client.

That feature has been an open issue for many years with respect to cpu and gpu resource control. Hasn't been worked on in a while by any existing devs as far as I know.

---