Climate Ocean Modeling on a Beowulf Class System.pdfVIP

Climate Ocean Modeling on a Beowulf Class System B.N. Cheng P. Wang Jet Propulsion Laboratory Jet Propulsion Laboratory California Institute of Technology California Institute of Technology Pasadena, CA, U.S.A. Pasadena, CA, U.S.A. Y. Chao M. Bondarenko Jet Propulsion Laboratory User Tech Associates, Inc. California Institute of Technology Pasadena, CA, U.S.A. Pasadena, CA, U.S.A. Abstract With the growing power and shrinking cost of personal computers, the availability of fast ethernet interconnec- tions, and public domain software packages, it is now possible to combine them to build desktop parallel computers (named Beowulf or PC clusters) at a fraction of what it would cost to buy systems of comparable power from supercomputer companies. This led us to build and assemble our own sys- tem, specifically for climate ocean model- ing. In this article, we present our experi- ence with such a system, discuss its network performance, and provide some performance comparison data with both R P SPP2000 and Cray T3E for an ocean model used in present-day oceanographic research. Keywords: Beowulf, cluster computing, ocean modeling, parallel programming 1 Introduction Beowulf class systems consisting of clusters of off-the-shelf PC’s are becoming a regular fix- tures in research and industrial computing. Traditional supercomputers are refrigerator- size cabinets that contain thousands of micro- processors. These supercomputers are built with specialized components and software that can be operated only by expert echnicians and programmers. These machines usually have their own cooling systems, require large amounts of electricity and cost typically more than $1,000,000. The Beowulf approach repre- sents a new business model for acquiring com- putational capabilities, particularly for small to medium sized applications. It comple- ments rather than competes wit