Resources, constraints, and job management

Points to make / outline

• impossibility of predicting exactly; a strategy
• how to measure with benchmarks, slurm, top (??); RSS as key thing to manage
• CPU utilization, context switching, overhead; threads, processes
• considerations for parallelism (perhaps also see parallel).

Standard resources: mem, disk, runtime, and tmpdir

Your own defined resources: other things.

Examples

• set up various memory constrained jobs and run with various different max memories; show overlap; make figure showing total memory used.