Queues and Policies

Queues

A user's job gets routed to a queue automatically, based on the specifications in the job.  There are 3 main queue, the normal queue (normal.q), the long queue (long.q) and the high priority queue (high.q).  Users DO NOT specify these queue resources.  Instead just specify the wall clock time you need.

Resources

Policies and Best Practices

• Jobs running longer than 12 hours will be routed to a limited pool of nodes for long jobs
• The high.c resource is intended for fast turn around jobs.  Users are limited to 2 slots for 12 hours.
• Always request the shortest wall clock time and the smallest amount of memory you need.  This will allow your job to get scheduled faster.
• There is currently a 2000 per user job limit.  If this impedes your workflow, please contact us.
• The long queue currently has the 1664 slots with the following distribution (Note all high memory nodes have the long queue enabled):
  160 48GB nodes => 1280 slots
  10 144GB nodes => 80 slots
  7 256GB nodes => 56 slots
  4 512GB nodes => 128 slots
  3 1TB nodes => 120 slots 

Current Shares

New Shares -- 

Each project may determine the subshares appropriate for their project.  For issues concerning share distribution, please talk to the proxy of the appropriate sponsor.

In order to run a job with one of the new projects and corresponding sub-projects, please add the following to your batch script (or just use -P if submitting to qsub from the command line):

#$ -P <user requested resource>

Where the user-requested resource is one of the sub-projects listed in the table below.