NEWS BRIEFS LIVEwire
Dallas, Texas –Supported by a grant from the National Science Foundation, The College of William and Mary, the nation’s second oldest college, has installed one of Virginia’s most powerful academic computer systems and the largest academic Sun Microsystems cluster in the Western Hemisphere.
Acquired by the College of William and Mary’s Computational Science Cluster, “SciClone” is a high-performance computer cluster with 160 processors and a peak performance of 115 billion mathematical operations per second. “A computational facility of this scale is somewhat unusual for an institution the size of William and Mary,” said Bob Voigt, director of the college’s Computational Science Cluster.
“SciClone represents a major increase in William and Mary’s academic computing capabilities. It will support faculty computing needs, graduate and undergraduate research activities and instructional activities across the sciences.” Comprised of computers with differing numbers of processors, differing speeds and different memory capacities, the heterogeneous system is organized as four subclusters that can operate as independent systems or in concert. It is this heterogeneous processing environment that makes SciClone one of the world’s most distinctive cluster computing systems and places SciClone at the forefront of cluster computing and computational science research.
“Most systems use the same technology for all communication, so this is fairly unique in the academic environment,” said Voigt. “Thus we have a rich testbed for conducting detailed studies of the interactions among algorithms, hardware and software systems.” Built entirely from Sun computers, SciClone uses three different networking technologies (Fast Ethernet, Gigabit Ethernet and Myrinet) and runs Sun’s Solaris operating system. SciClone’s unique computational environment allows researchers to study effective use of heterogeneous collections of processors located throughout the Internet. This concept of grid computing – where researchers, regardless of their physical locations, can access many different kinds of systems-actually was one of the guiding principles in the design of SciClone.
“We purposely created a controlled, grid-like environment that is complex enough to allow faculty and students to research those issues facing users of larger distributed systems,” said Tom Crockett, senior research associate at William and Mary’s Computational Science Cluster, and principal architect of the system.
For more information about SciClone, access http://www.compsci.wm.edu/SciClone/