University at Buffalo - The State University of New York

Published: November 21, 2002

The university has added a 300-node Dell high-performance computing cluster (HPCC) to its Center for Computational Research (CCR).

The increased computing capacity will assist with various scientific research projects, including groundwater modeling to help predict the flow of contaminants in large bodies of water such as the Great Lakes, computational chemistry and molecular structure determination.

It is the second Dell cluster at UB, adding to the 2,000-node HPCC deployed earlier this year to support research in the Center of Excellence in Bioinformatics.

CCR, the eighth-largest supercomputing site in the world, underscores how standards-based computing systems can perform at high levels for complex research. The 300-node cluster recently achieved 2.004 trillion floating-point operations per second (Teraflops) of sustained performance in the LINPACK benchmark test.

"Many of our scientists need to exploit a large number of processors operating in a coordinated fashion to jointly solve leading-edge scientific problems that could not be solved in a reasonable amount of time on smaller systems," said Russ Miller, director of CCR and UB Distinguished Professor in the Department of Computer Science and Engineering. "Efficient massively parallel processing techniques can be applied to many scientific problems in order to provide cost-effective solutions via clusters based on standard components. A machine like the Pentium 4-based Dell cluster will be used to dramatically reduce the time to solve problems, in many cases from months to hours."

UB and many other organizations are increasingly choosing HPCC solutions for data-intensive analysis as an alternative to proprietary supercomputers. The latest Top 500 List of supercomputers www.top500.org indicates that Dell clusters have a cumulative performance of 6.046 TFLOPS, up from 856 Gigaflops (GFLOPS) in the previous list.

The new supercomputing cluster at UB is the highest-ranking Dell system on the list at number 22. Other ranking Dell clusters include Sandia National Labs (32), Cornell Theory Center (88), University of Utah (89), Penn State University (174), Swineburne University (180), a 100-node configuration of UB's first cluster (187), Dell (207) and the University of Notre Dame (461).

A key reason for the popularity of HPCC for supercomputing applications is the ability to deploy solutions based on standardized technologies at a fraction of the cost of a proprietary supercomputer.

"Dell's supercomputing clusters enable customers to scale as they grow, as opposed to paying for overcapacity in the legacy supercomputing model," said Russ Holt, vice president and general manager of Dell's Enterprise Systems Group. "The ability to add power and capacity to address demand is very attractive to customers, and very cost-effective."

UB's second cluster is comprised of 300 Dell PowerEdge™ 2650 servers, each with dual Intel® Xeon™ Pentium 4, 2.4 GHz processors running Red Hat Linux. A fully connected Myrinet 2000 high-speed, low latency interconnect network completes the balanced standards-based supercomputer.

CCR researchers will use the cluster for work ranging from groundwater modeling, protein folding, molecular structure determination and computational chemistry to environmental engineering, computational fluid dynamics and materials science.

Dell Computer Corporation is a premier provider of products and services required for customers worldwide to build their information-technology and Internet infrastructures. The company's revenue for the past four quarters totaled $33.7 billion. Dell, through its direct business model, designs, manufactures and customizes products and services to customer requirements, and offers an extensive selection of software and peripherals. Information on Dell and its products can be obtained at www.dell.com.

Editor's Note: This story was based on a news release written by Dell Computer Corporation.