The support resources have allowed for us to bring three new head nodes online for cluster wide usage. These new head nodes offer failover capability as well as more resources, at 12 cores and 128GB of memory each. The new head nodes utilize a Constellation naming scheme, and are accessible by logging into a singular Round-Robin alias: "Orion". The alias is simply a name that rotates connections to each node, and is intended to reduce overloading of a singular head node, as commonly occurred on rhel6head1 in the past. Each head node can be logged in directly under their true names - Hydra, Libra, and Virgo; but the primary recommended name to connect to is "Orion". This means that our older head nodes, rhel6head1-4, will be retired and unavailable for login come March 1.
With the Spring semester now in full swing, we hope that everyone is doing well. At the end of last semester, the HPC team purchased new nodes to support new support services as well as additional compute capacity to the cluster. The additional compute capacity comes in the form of six new compute nodes, currently in their final testing phase, and once open will add an additional 264 cores to the cluster. This brings our total core count to 4080 cores!
One last change is the addition of a new hardware specific SLURM partition: onegig. As a means of waning unneeded use of our tengig nodes, the "onegig" partition only contains our older, non-infiniband bladecenter nodes. The benefit of using this partition being that it has a maximum priority value, higher than our small/medium/large partitions, allowing specified jobs to run more quickly. For those users who simply need cores regardless of system type, the idea here is to reward not using highly demanded tengig nodes with a shorter wait time.