I run condor on a small academic cluster. The system consists of about 9 machines, each with 4-8 cores. Devoting one machine each to the submit and central manager node roles seems a resource inefficiency (ie, about 20% of the available
threads and power consumption goes to submit and manager role). I realize this is an edge case for most condor installations, but in small systems like mine is it possible to have the central manager and/or submit machines also serve as execute nodes? Nathan Moore Winona State University |