High-Performance Computing Cluster ⟶
I assisted in designing and constructing a Beowulf-style compute cluster for PrISUm out of recycled hardware. The cluster consisted of seven desktops with quad-core Intel Core i7 processors with 8GB of RAM, each running a build of CentOS. This gave the cluster a combined compute capacity of 28 CPU cores and 56GB of shared memory. All of the nodes were connected on a local gigabit ethernet network. The head node of the cluster ran Sun Grid Engine for job scheduling and had 1TB of network storage for applications and user workspace. The cluster was primarily used by the team for computational fluid dynamics simulations with STAR-CCM+ and MATLAB optimization problems. When the cluster was not in use, it contributed to community grid computing projects.
This is how I originally found the cluster when no one on the team knew how it worked or how to use it:
This is after we (myself and Michael Wrona) were able to tear down the server rack completely, and refurbish and rebuild all the computers: