ideal resubmission frequency on ARCHER
|Reported by:||fpithan||Owned by:||um_support|
|Platform:||ARCHER||UM Version:||<select version>|
I would like to reduce the time my jobs spend in ARCHER queues to speed up the experiment's total turnover. I am running the UM version 8.5 at N96, AMIP-type, using 96 cores and 12 cores per node. My current resubmission pattern is once per year. Would it be better to resubmit less frequently to exploit the maximum runtime per job, or more frequently to have shorter individual jobs? Or is one year already the ideal balance?