Compute infrastructure consists of a pair of older servers for compute-intensive tasks. The nodes are as follows: 2 of (2x intel Xeon E5-2698v3, 128GB of ram) with a 10Gbe interconnect. These nodes are nominally superior to those used in the "wheeler" cluster used at the University of New Mexico, whilst being about a generation behind the latest clusters used at ComputeCanada's Cedar cluster. All of them run ubuntu 22.04 as the operating system, and use slurm + openMPI as the stack. They have pre-installed CP2K, psi4, terachem, and Orca, openFOAM and WRF. Sadly as I dont have a personal Gaussian license, I can't use that. Precompiled software from compute canada (see the wiki) is also available
Storage for the cluster is via a NFS share, backed onto a unbacked-up ext4 filesystem. I don't have the funds to spring for more drives right now for RAID, so I'm just backing up my data to my instituitionally provided cloud backup. There is also no data on the cluster that is of production importance.
Energy usage is a concern, at full throttle each server consumes approximately 300 watts as measured from the wall. Thankfully however, Quebec has the cheapest electricity in north america, and I don't pay for it in my rent as long as total energy consumption is below my rental contract's limits, which I don't reach