The example provided in the slides may be too specific. Consider adding higher-level examples about how to set up and start a workflow manager and where persistence1 comes in. Michael's suggestion for a possible set of diagrams:
Login Node > Launch Manager > Manager runs Slurm jobs (BAD!)
Compute Node > Launch Manager > Manager runs Slurm jobs (Ok, but restricted to 24 hr or 7 days. Must also wait for resources to start - not ideal for long running workflows).
Persistence1 > Launch Manager > Manager runs Slurm jobs (Great!)
The example provided in the slides may be too specific. Consider adding higher-level examples about how to set up and start a workflow manager and where persistence1 comes in. Michael's suggestion for a possible set of diagrams:
Login Node > Launch Manager > Manager runs Slurm jobs (BAD!)
Compute Node > Launch Manager > Manager runs Slurm jobs (Ok, but restricted to 24 hr or 7 days. Must also wait for resources to start - not ideal for long running workflows).
Persistence1 > Launch Manager > Manager runs Slurm jobs (Great!)