Demetrios gave an extensive talk about the uses of cluster computing, including three live demonstrations. The beginning of the meeting was an overview of the Bootable Cluster CD (BCCD) project. Demetrios went over the advantages and disadvantages of BCCD. It’s a way of creating a cluster of computers without needing to make the setup permanent. We all moved into a lab after the introduction and created a BCCD cluster with a few CDs. We were able to witness the performance increases that come with parallelizing software across a BCCD cluster. Specifically, we used a program named “GalaxSee”, an n-body simulator, to test the ability of the cluster.
After the BCCD lab, Demetrios took us to a lab at Stockton containing 5 Nvidia Tesla cards. The whole group took turns exploring some of the basic example software made available by Nvidia. Demetrios described CUDA, an engine used for interacting the the GPUs on the Tesla cards. The GPUs on the Tesla cards allow researchers to perform calculations at a faster rate than current CPUs and at a lesser cost.
Demetrios finished his presentation by demonstrating a Rocks cluster he built. The Rocks cluster comes with several software packages that help the administrator monitor and use the cluster. Ganglia is included with the Rocks cluster and gives the administrator a scalable distributed monitoring system. Slides from this presentation are available!