Skip Navigation

University of Nebraska–Lincoln

Research Computing Facility

Homestead Miser

Miser Batch Use

  • Queues Available -- There is presently one queue: mpp.
  • mpp: 24 cpu, 4.5 GB RAM All users running parallel jobs may access this queue. Users are asked to limit jobs exceeding 1 hour (wall-clock time) to 8 cpu maximum, and 24 hours/cpu maximum runtime. Special arrangements for more cpu at one time, or longer run times, may be made upon request.
  • Submitting Jobs
    • Determine the size of your job. Miser will allocate the amount of memory, number of cpu's, and total time (run time * number cpu) for your job. Please try to accurately estimate these numbers. Users are stongly encouraged to checkpoint long jobs. This may ther be done by outputting intermediate files, or by using the SGI routine cpr (there is a GUI cview for this purpose).
    • Make sure your job adheres to maximum job limits.
    • Submit your job.
    • Example: for an executable 'run' that requires 2 cpu for 24 hours (each) and 100 MB of memory (total), the miser command looks like

      /usr/sbin/miser_submit -q mpp -o t=48h,c=2,m=100m run &

      This example uses the integer-unit specifier format for time. all miser commands are located in /usr/sbin Job limits per cpu (in mpp) are now 24 hours, so multiply accordingly when submitting multiple cpu jobs.


  • Monitoring Jobs The bid mentioned in this link is simply the process id (found by using the ps command).
  • Deleting Jobs -- users are asked to use miser_kill. Simply killing a miser job with the 'kill' command can have negative side effects.
  • Caution: the environment variable MP_SET_NUMTHREADS is normally not set, but if it is, it may conflict with the number of cpu asked for in miser, leading to unexpected results. If miser does not grant you the expected number of cpu, setting this to the number of desired cpu will fix the problem. For the example above, one could use setenv MP_SET_NUMTHREADS 2 to 'force' miser to behave. (MP_SET_NUMTHREADS may be used to vary the number of cpu a parallelized code will potentially grab interactively).
  • Caution: miser is tempermental ... if you have trouble, please contact RCF staff.