Home > Research Support Group > WestGrid > Local Usage Policies

Serial/Parallel use:

• The Origin machines are expected to be used to run large parallel jobs which cannot be run elsewhere.
• For Serial jobs and for jobs requiring distributed memory, WestGrid has a large Compaq Alpha cluster in Calgary, and a large Linux cluster at the University of British Columbia for its members. Check the WestGrid web site for more information.
• While the SGI Origins can be used for parallel work, WestGrid is encouraging all of its members to use the Compaq machines for any serial, i.e. single processor, problems that members have to run.
• In addition AICT maintains machines specifically for scalar numerical work, the numerical server, as well as a Linux cluster which provides a parallel environment. Please contact Research.Support@ualberta.ca if you wish to pursue the move of your jobs to these other machines.
  
• Nexus, an 8 processor SGI Origin model 350:
  • To be used for interactive and batch processing.
  • Interactive jobs:
    • must not use more than two CPU's, and
    • must not use more than 60 minutes of walltime.
  • Short interactive jobs are intended to allow users to compile and test their programs. This is to reserve most of the machine for large batch jobs. Note that interactive usage is monitored and clients who repeatedly exceed 60 minutes of walltime may have their processes killed.
    
• Arcturus, Aurora, Borealis, Australis, Corona and Helios:
  • There is no interactive use on these machines.
  • All processors are reserved for batch processing.
  • Only parallel jobs using 64 or more cpus can be submitted to Arcturus.
  • Jobs that require 64GB of memory or more can also be submitted to Arcturus.
  • Programs to be run on Arcturus may be evaluated by Research Support to ensure efficient use of the machine.

Batch Scheduling Policy:

• The Portable Batch System (OpenPBS) and the Maui Scheduler are used to manage all non-interactive jobs running on Nexus and all jobs running on the other UofA WestGrid machines.
• PBS keeps track of jobs submitted and runs them in order of their priority.
• All clients are expected to submit their jobs to PBS/Maui rather than running them directly, except for limited interactive use permitted on Nexus.
  

PBS usage rules for the queues:

• Jobs will be run using a "fair share" priority system, which calculates priority based on the amount of recent CPU usage by that user/group. The priority formula depends on the amount of recent usage.
• For Jobs that require a "large" amount of memory (i.e., more that just a few hundred MBytes), please request enough CPUs to cover your memory usage. We have 1 GB per CPU on Arcturus. Defaults for memory and time are set to encourage inclusion of these resource requests; however users that have very high memory usage and low CPU count may need to request more than 64 CPU's.
  

Walltime and Checkpointing:

• The maximum walltime (physical elapsed time) for jobs on Arcturus, Aurora, Borealis, Australis, Corona and Helios; is 24 hours.
• Please use checkpointing (saving the current state of your program before a restart) to avoid losing the results of your calculations during a restart.
  

Disk Usage and Backups

• The SGI Origins share disk space mounted on an SGI TP9400 and an SGI TP9500 disk array.
• Clients must be aware that as with all file system and back up facilities, problems can still occur. Users should still insure that their critical files are archived in a separate location.

If you have concerns about this policy, please contact us at: Research.Support@ualberta.ca.

The research support team is available to assist users in porting code and ensuring that their programs are making efficient use of the parallel architecture of the SGI Origins.

Revised: Jan. 14, 2005