DWalker487
commented
5 years ago Thanks for this, (sorry just got back from some time away). It all looks good by eye.
I implemented the splitting on advice from Adam, as we can end up inadvertently swamping one of the computing elements if only one is used. This was a while back so standard procedure may have changed. I guess the most optimal solution would be assigning ce with a probability based on the relative speeds of ce1 and ce2, but that's probably overkill for now :P
Added -S flag for direct submission (direct selection of protocol to use, which should save time) Don't arbitrarily swap between ce1 and ce2, use only as directed (ce1 is faster than ce2, ce2 really is the backup ce to use in case of failure of ce1).