What HPC facilities should I use, Stern or NYU
In general, you should probably start with the Stern facilities if you are unfamiliar with HPC processing, or don’t have huge processing needs. We can give you hands-on support to get you going.
Generally, the Stern HPC has faster turnaround than the NYU HPC, but that can vary. The NYU HPC facility supports the whole global university and has significant capabilities, including supporting a larger collection of software than Stern. In the Fall of 2020, the NYU HPC a new system (Now GREENE.HPC.NYU.EDU) was installed that is has 30,000 cores of processing, and a large number of GPUs. It is one of the most powerful HPC clusters in academia. Our staff can help you decide where it is best to do your processing.
However, currently, Stern currently uses a different scheduling system than the NYU HPC (although we are transitioning to Slurm). We currently (Winter 2021) support both grid engine and slurm. Slurm is a more powerful scheduling system, which supports (among other things) scheduling jobs on GPUs. For that reason, we are migrating to slurm, so that users can test their code on the Stern HPC cluster, and, if necessary, migrate to the NYU HPC.