I definitely like Ubuntu for workstations myself, although every cluster
I've worked on has used CentOS.
For the server side I've tried both SLURM and ROCKS and I would go with
SLURM myself. It seems to create fewer headaches with installing new
software and packages and making them available to compute nodes.
For getting GPU up and running, SLURM seemed much more straightforward as
well, whereas ROCKS didn't seem to have as much built in support ...
although that was a couple years ago and I haven't checked out the newest
version of ROCKS, so maybe that has changed.
I personally like the CentOS - SLURM combination myself. I've been able to
get AMBER15, including a CUDA-MPI capable configuration, up and running on
it with (relatively) little fuss.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri May 20 2016 - 16:00:04 PDT
