Iupui Mason High Performance Mass

In the iupui mason server

  • there are 20 server, each have 10 gigabit ethernet
  • 4 cpu in each server
    • 8 core processor each cpu
    • total 32 core in each server
  • in the /scratch or /tmp there are total 400 GB disk storage
    • and the file will be delete after 14 days since they are created

when login

cpu/memory limit

  • user process on login node are limited to 20 minutes CPU time
    • if exceeded, will be killed automatically
  • use TORQUE qsub to submit job that need more than 20 minutes cpu time
    • qsub

      • qsub -l walltime=10:00:00 job.script
        • for the system default time is 60 minutes
          • so use walltime to use more time
      • qsub -l nodes=1:ppn=2 job.script
        • run the script in a node and using two core processers
      • qsub -l nodes=4:ppn=31,vmem=100gb -l walltime=20:00:00 jobscript.script
        • if use multi commands, use -l to separate the argument
    • qstat

      • qstat -a
        • display all job
      • qstat -n
        • list the nodes allocated to a job
      • qstat -r
        • list the job that are running
      • qstat -u username@host
        • display the job owned by the username
    • qdel

Comments