Already in 2003 when we started the Westmont Cluster Project, it was clear that the future of high-performance computing would involve the exploitation of massively parallel architectures and programs. The simplest and cheapest approach constructs clusters of generic PC processors tightly linked to provide intensive computational capabilities. In 2003, such clusters were sometimes known as Beowulf Clusters. We have been using the Rocks Cluster Distribution during the life of our project.
We initially conceived the Westmont Cluster Project as a service opportunity with two primary benefits. First, the people working on the project would gain experience constructing, configuring, administering and programming a cluster system. While we do not anticipate that most participants will be studying parallel architectures or programming in graduate school nor even necessarily working with super-computers in the commercial sector, we are confident that their experience with a cluster system will prove valuable either in graduate school or in industry. Beyond the administrative experience with the cluster, the platform will allow participants to study and develop existing and novel parallel algorithms. Such algorithms are increasingly important in the current software development environment.
As the second primary benefit arising from the Westmont Cluster, we can support the computational needs of multiple client users. During the course of the cluster project, four different research efforts at Westmont have relied extensively on the cluster; three of those research projects are ongoing as of 2013. Between 2004 and 2006, we also provided computational services for a research project based at UCSB's Chemistry department. Two new client projects are currently under consideration.