HPC Queues

In order to provide fair access to compute resources in the HPC Compute Cluster, we employ the Moab Job Manager. Thus, when you wish to run a job in the HPC, you must submit that job to a queue to be processed in the order in which it was submitted.

General Access Queues

Everyone with an RCC Account has access to four queues:

Queue Queue Name Walltime Clock Limit Walltime Clock Default Max Jobs per User Max Processors Per User Max Processors
General Access genacc_q 14 days 14 days 4/8* 608 608
Backfill backfill 4 hours 4 hours 100 512 n/a**
Backfill 2 backfill2 4 hours 4 hours 100 512 n/a**

* There is a limit of four jobs in the genacc_q. However, if slots are available, users can run up to eight jobs
** Backfill queue jobs can utilize most all nodes in our system

The General Acccess Queue is the queue most users will use to submit long-running jobs (up to 14 days).

The Backfill Queue provides access to more resources, but only allows jobs to run for up to 4 hours. After this time limit, the system will automatically kill the job. The purpose of the backfill queue is to allow users to run short-running jobs on larger resource sets to fill gaps in the processing schedule. For example, if the system determines that a number of processing nodes will be idle for a given period of time, it can insert a job from the backfill queue to use those resources.

Queue Parameters

  • Queue Name - The name of the queue in the system.
  • Work Clock Limit - The maximum amount of time a single job is allowed to run. After this time limit has passed, our system will automatically kill the job if it has not already completed.
  • Work Clock Default - The default amount of time a job is set to run for. This is specified by the user when the job is submitted, and can be any value less than or equal to the Work Clock Limit. This value should be chosen carefully. If you choose a value too low, the system may kill your job before it has completed. If you choose a number too high, the system may not run your job as quickly, due to scheduling rules.
  • Max Jobs Per User - This refers to the maximum number of jobs that can run simultaneously for any particular user.
  • Max Processors Per User - This refers to the maximum number of processors a single user can use simultaneously to process jobs.
  • Max Processors - This refers to the maximum number of processors that all jobs in this queue can simultaneously use in aggregate. If three jobs are running, for example, they may not use more than the maximum number of processors.

Owner-Based Queues

If your account has access to owner-based queues, please refer to the information on your Account Page for details about which queues you have access to.