High Performance Computing
High Performance Computing (HPC) is the use of advanced computing software and hardware to reduce the time needed for complex computations. Traditionally, HPC depended on supercomputers: highly fast and/or parallel systems that cost millions of dollars to own and operate.
In the early 1990's, improvements in processors and memory, networking hardware, and open-source software led to the development of new, cost-effective alternatives for HPC, including grid-computing, clustering, and cycle-scavenging.
ETSU's HPC Project
In January 2005, the CSCI department commissioned a series of graduate-level capstone projects to develop cost-effective strategies for implementing HPC in the ETSU environment.
- The first project, which ran from January 2005 through May 2006, reviewed public-domain HPC software packages, parallelized two research simulations of interstellar phenomena for ETSU's physics department, and prototyped a Condor-based system for parallelizing the rendering of digital images for ETSU's Digital Media program. Preliminary results showed that all three tasks run appreciably faster in an HPC environment.
- The second project, which ran through May 2007, successfully deployed Dr. Beverly Smith's simulation of galactic collisions, and determined that Condor on Windows was too unstable of a platform for supporting parallel rendering in the ETSU environment.
- The third and final project, which will run through May 2008, has been tasked with tuning the performance of Dr. Smith's software and investigating the use of BOINC for parallel rendering. The team will also be responsible for helping ETSU to document and investigate uses for its new, 256-node high performance computer.
About This Website
This website contains case studies, code, and other supporting documents created by the HPC Project Teams over the course of this project.
