Running jobs on midway

This section of the RCC User Guide is about running computations on the RCC cluster, Midway2. All jobs running on compute nodes consume Service Units (SUs); see RCC Service Units for more information.

Login nodes vs. compute nodes

Once you have connected to Midway2 (see Connecting to RCC Resources), you may work on one of the login nodes. Login nodes may be used for compiling and debugging code, installing software, editing and managing files, submitting jobs, or any other work that is not long-running or computationally intensive. Login nodes should not be used for computionally intensive work.

All intensive computations should be performed on compute nodes. Access to compute nodes can be obtained by submitting a job through the Slurm scheduler, or by requesting an interactive session (see Using Midway). If you are unsure whether your computations will be intensive, please request an interactive session and continue your work once you have connected to the compute node.

Note

Running computationally intensive jobs on the Midway2 login nodes prevents other users from efficiently using the cluster. RCC System Administrators may terminate your processes without warning if your processes disrupt other users’ work on the RCC cluster.

For more information on how to interact with Midway2 through the Slurm resource manager, see Using Midway.

Illustrative examples

Below are a number of example submission scripts that you can adapt to run your jobs on Midway2. For information on how to run specific software packages on Midway2, consult the Software section.