It would be nice (and will probably be necessary on MSI in the future) to be
able to resume running GNU parallel after it has been killed/stopped by the
scheduler because it did not finish everything. Not in the case when # allocated
nodes = # simulations to run, but in the case when # allocated nodes < #
simulations to run, and GNU parallel steps through the commands file
incrementally. In this case, being able to resume a previously "suspended"/killed
simulation run will be very valuable.
And necessary if I'm going to be getting tiny bits of time/nodes on the new
128-core cluster! Definitely not going be able to request 50 nodes
simultaneously on there...
It would be nice (and will probably be necessary on MSI in the future) to be able to resume running GNU parallel after it has been killed/stopped by the scheduler because it did not finish everything. Not in the case when # allocated nodes = # simulations to run, but in the case when # allocated nodes < # simulations to run, and GNU parallel steps through the commands file incrementally. In this case, being able to resume a previously "suspended"/killed simulation run will be very valuable.
And necessary if I'm going to be getting tiny bits of time/nodes on the new 128-core cluster! Definitely not going be able to request 50 nodes simultaneously on there...