There are two sets of methods of running programs and jobs on the HBS compute grid. The easiest, using default submission scripts via NoMachine and pre-defined commands, was covered in the previous section. Here we describe using custom submission scripts, which offers maximum flexibility but does involve writing one-line unix commands or simple, text-based submission scripts.
Using custom submission scripts to run interactive or batch jobs offers you the full power of the grid – you specify the resources you require that for your work in order describe the shape of your job to the scheduler. You are only limited by the size of the cluster and any system-wide enforced limits. As described previously, these resources include:
- Amount of RAM
- Total run time for your analysis
- Number of processors
- Which set of compute nodes the job should run on
If you do not specify any of these options, your job will be run with the system default, which is typically not what you want.
Furthermore, your job submission can be done on a single command line, which is perfect for running a single command as a batch job; or as options in a text file, which has a list of commands that you wish to perform as a part of your job. And we'll dive into these in the next sections: