Most Cheyenne batch queues are for exclusive use, so jobs submitted to those queues are charged for all 36 cores on each node that is used. Jobs in the shared use "share" queue are charged only for the cores that are used.
The "regular" queue, which has a 12-hour wall-clock limit, meets most users' needs for running batch jobs on Cheyenne.
Cheyenne queue details
Some additional queues on the system are for dedicated purposes and accessible only to authorized users. Casper: See this page for information on running jobs on Casper nodes. Calculating chargesExclusive nodesCharges for use of Cheyenne are calculated in terms of core-hours. Jobs run in Cheyenne queues other than "share" are charged for exclusive use of the nodes by this formula: wall-clock hours × nodes used × cores per node × queue factor Number of nodes usedYour batch script indicates how many Cheyenne nodes your job will use. In this example, you have selected 2 nodes, each of which has 36 cores. Your job will be charged for the use of 72 cores.
Shared nodes (Cheyenne and Casper)Charges for jobs that you run on a shared node, including Casper nodes, are calculated by this formula: core-seconds/3600 (core-hours) Checking and managing chargesUsers can check computing and storage charges through the CISL Systems Accounting Manager. (Go to documentation or to SAM app.) If you have concerns about using your allocation most efficiently, contact the NCAR Research Computing help desk for guidance. Sometimes jobs can be configured to make better use of the processors, and you may be able to save by using a less expensive queue. CISL can refund core-hours if system failures cause jobs to fail and the failed jobs are reported promptly. Use this core-hours refund request form (login required) if you think a refund is warranted. Technical limitations prevent us from verifying refund eligibility for jobs that are more than seven days old. |