...
- automatic job suspend/resume ("idle queue" behavior) (really want)
- automatic afs tokens for mpi jobs on multiple nodes (would be nice)
- job environment propagation (including limits) (necessary)
- subgroup-specific accounting (control of queue-specific priority formula) (really want)
- subgroup-specific administration (e.g. subgroup administrator btop, bkill privilege) (necessary)
- cross-queue fairshare (with cpu-time weight) (necessary)
- cpu reservations for MPI (necessary)
- gpu support (necessary)
- ability to submit jobs on hosts where we don't have accounts/home-directories (really want)
Advisory
...