Computational Research Center
The University of North Dakota (UND) Computational Research Center (CRC) is dedicated to helping researchers affiliated with the University of North Dakota solve increasingly challenging problems in science and society.
By supporting high performance computing hardware, software and staff through the Division of Research, the University provides quality computing resources to its researchers and faculty involved in creative activities which require intricate numerical modeling and data-intensive simulations.
Weekly Maintenance Notice
Routine maintenance is scheduled for the second Thursday of every month.
We have discovered a problem with old PuTTY versions causing login issues and/or performance degradation. Please ensure that your PuTTY version is up-to-date. The current PuTTY version is 0.76.
The Globus server is experiencing multiple issues. The first issue is that the storage mount is out-of-sync. The second issue is that group permissions are not working. We are in the process of deploying a new Globus server which will resolve both of these issues.
Effective immediately, the Hodor and Arya clusters will be queues under the new Talon cluster. You will now log in to talon.UND.edu for all job submissions. Hodor is becoming two queues, one for GPU jobs (hodor-gpu) and one for CPU jobs (hodor-cpu). To submit jobs to Hodor or Arya, simply add the following line to your submission script:
queue_name will be the queue you want to use. If you don't specify a queue, it will default to talon which is the default queue. Below is the list of available queues:
hoffmann - Arya (authorized departments only)
manu - Arya (authorized departments only)
hodor-cpu - Hodor CPU
hodor-gpu - Hodor GPU
talon-gpu - Talon GPU deep learning nodes (authorized users only)
talon-fat - Large memory
talon - Talon CPU, this is the default queue
For scheduling fairness and backfill scheduling efficiency, all Talon and Hodor queues now have default and maximum wall times enforced. The default and maximum wall times can be found below. The maximum wall times can also be found by running sinfo. Maximum wall times are subject to maintenance windows. If the default 2-hour wall time is not enough, then you will need to set your wall time in your submission script. Application checkpointing is always strongly encouraged.
hodor-cpu - Default run time, 2 hours. Maximum run time, 8 days.
hodor-gpu - Default run time, 2 hours. Maximum run time, 8 days.
talon - Default run time, 2 hours. Maximum run time, 28 days.
talon-fat - Default run time, 2 hours. Maximum run time, unlimited.
talon-gpu - Default run time, 2 hours. Maximum run time, 28 days.
How to Acknowledge UND CRC in Published Works
“This work used advanced cyberinfrastructure resources provided by the University of North Dakota Computational Research Center.”