How to Run Codes—Your Own or Commercial/Open Software
Whether you run your own code, e.g., built using some programming language, or a commercial/open source code such as Ansys, you will be interacting with a cluster’s job scheduler. Our scheduler on ARC clusters is the Slurm Scheduler. The first video is an excellent and concise introduction for running interactive and batch jobs. The subsequent videos are motivated by the recurring need to be more nuanced in specifying interactive and batch jobs, and hence these give examples on different classes of interactive and batch jobs that you will use.
The basics of how to run both interactive and batch jobs.
Learn the different types of compute nodes on clusters and how to specify them for your jobs. Watch this before the later videos of this section.
A more detailed example of an interactive job that uses the material in the second video—that is, how to use constraints to specify compute node types.
A more detailed example of a batch job that uses the material in the second video—that is, how to use constraints to specify compute node types.
Examples of slurm batch jobs where files (input files, code files, output files) use volatile resources (in particular, volatile storage) to speed file input/output during code execution. The volative storage resources are referred to as localscratch and are discussed more here. The two examples in this video illustrate the use of
$TMPFS(use of a compute node’s volatile memory) and$TMPDIR(use of a compute node’s hard drive).