2

On a PBS system I have access to, I'm running some jobs using the -W x=NACCESSPOLICY:SINGLEJOB flag and, anecdotally, it seems that the same jobs take about 10% longer when adding this flag as without. Is this correct behaviour? If so, it surprises me, as I'd have thought having sole access to a whole node would, if anything, slightly decrease the time taken for a job to run due to having access to more memory.

clusterdude
  • 616
  • 3
  • 16
ChrisW
  • 4,970
  • 7
  • 55
  • 92
  • Are you counting the time that it takes to get scheduled and then run, or just the run time? If you're looking at the total time including getting scheduled, then it would make perfect sense, as there may be scheduler policies that delay jobs that are requesting exclusive node access. – dbeer Dec 14 '11 at 22:47
  • No - I do mean just run time. (The queue time is also longer, but I was obviously expecting this) – ChrisW Dec 15 '11 at 12:48
  • 1
    I would check if the scheduler has policies that restrict singlejob access to certain nodes, and if those are weaker hardware. Also, how many times have you experienced this? If it isn't very many times, there could be pseudo-random factors that affect it. – dbeer Dec 15 '11 at 16:32
  • Thanks - I hadn't thought of that, could certainly be a factor. – ChrisW Dec 21 '11 at 22:13

0 Answers0