Jobscript Generator

Build a Slurm batch script for the Lawrencium CPU and Einsteinium GPU clusters, and see roughly how many Service Units (SUs) the job will use. Choose a partition, set the resources and time limit, then copy or download the script and submit it with sbatch. Your choices are saved in this browser. To share a setup with someone, use Copy link to this setup.

How the Estimate Works

The estimate follows the Recharge Model:

SUs = SU ratio of the partition × CPU cores allocated per node × number of nodes × hours

• Shared partitions (lr7, lr8, cm1 and the GPU partitions) charge for the cores the job requests: tasks per node × CPUs per task. On GPU partitions, the number of cores is set by the number of GPUs, since each GPU must be requested with a fixed number of CPU cores. For example, es2 needs 14 cores per GPU, which at an SU ratio of 4.0 comes to 56 SUs per GPU-hour.
• Exclusive partitions (lr4, lr5, lr6 and lr_bigmem) allocate whole nodes, so every core on the node is charged even if the job uses fewer. When a partition has nodes with different numbers of cores, the estimate is shown as a range.
• The estimate uses the time limit you request. Charges are based on how long the job actually runs, so a job that finishes early uses fewer SUs than estimated.

What the SUs cost depends on the project account the job runs under:

Every user account is also charged a $25/month account fee, whether or not jobs are run. See Charges and Billing.

sacctmgr show association -p user=$USER

The generator covers the main Lawrencium and Einsteinium partitions. For other partitions, such as cm2 and cf1, and for the supported research clusters, see Example Scripts and the Computing Systems pages.