Date Posted: 4/01/2000

ITHACA, N.Y. - The Cornell Theory Center (CTC) announced today acquisition of a new 128-processor, high-performance cluster to meet a growing demand for large-scale parallel computing resources. The new Velocity+ cluster, funded through CTC's Advanced Cluster Computing Consortium (AC3), is the latest Windows-based Dell/Intel/Microsoft cluster to be installed at CTC. Velocity+ will be dedicated to strategic, computationally challenging applications aimed at solving highly complex problems in fields such as molecular modeling and multiscale materials simulation.

The need for additional resources for large-scale applications at CTC became apparent almost immediately after Velocity came into full production in March. "The user migration to Velocity has proven very successful," said CTC executive director Linda Callahan. "Many parallel users are reporting extremely good performance and scalability."

"You might say we quickly became victims of our own success," added CTC director Thomas F. Coleman. As new parallel users scaled up to larger runs, overall demand for Velocity made it difficult to provide sufficient resources for the key research areas of protein folding and multiscale materials that are highly parallel, large-scale applications and challenge the limits of any system."

Partners Dell, Intel, Microsoft, and Giganet worked with CTC to provide the required capabilities and to demonstrate the effectiveness of industry-standard hardware and software for the most complex high-performance computing problems. We are extremely excited to be part of this ambitious effort, said Scott O'Hare, Dell vice president and general manager of the company's higher education segment. Seeing Dell's PowerEdge servers capable of meeting the scaling, reliability and performance requirements of the new Velocity+ cluster -- and contributing to such groundbreaking research -- is rewarding to all of us at Dell. We appreciate the confidence that the CTC team has shown in Dell."

Velocity+ consists of 64 dual Pentium III 733 Mhz processors, with 2 Gigabytes RAM per node, 27 Gigabytes of disk space (RAID 0), and full 64-way Giganet Interconnect. Velocity+ brings to 472 the total number of processors in the Velocity complex. The system runs Microsoft Windows 2000, which is key to the seamless desktop-to-HPC model that CTC is pursuing.

"Microsoft is pleased to support the Velocity+ effort," stated George Spix, director of Advanced Systems Enterprise Server Systems at Microsoft. "Cornell is continuing its 30-year tradition of bringing leading-edge computing to their community and partners." Spix continued, "Especially encouraging is Cornell's focus on real-world HPC commercial, scientific and engineering, and biomedical applications, and their seamless integration with desktop and departmental engines."

One of the strategic applications lined up to run on Velocity+ comes from CTC's Parallel Processing Resource for Biomedical Scientists, which is funded by the National Center for Research Resources. Ron Elber, lead scientist for the Resource has already seen exciting results in his new stochastic path approach for modeling protein dynamics applied to hemoglobin. His group has captured the essential dynamics of the hemoglobin switching process using STO (part of his MOIL package) on up to 96 processors with near linear speedup. Elber's software has been optimized for Windows-based systems. The packages are made available free to the scientific community. World-renowned scientist, Harold A. Scheraga will also have a strategic application running on Velocity+. Scheraga's group conducts groundbreaking research in the ab initio solution of protein folding, applying fundamental physics -- first principles -- to solve the structures of proteins.

CTC's Computational Materials Institute (CMI) focuses on multi-scale collaborations across disciplines and institutions aimed at the design of algorithms and systems for multi-scale engineering - the numerical simulation of crack propagation, for example. According to CTC associate director and CMI lead researcher, Anthony I. Ingraffea, "One of CMI's goals is to provide multiscale computing on demand from the desktop - which means that researchers can predict materials performance at the atomic level all the way up to the product level."

CMI researchers collaborate with a number of corporations, including Alcan, Allegheny-Ludlum, Boeing, Eastman Kodak, and Northrop Grumman, and are supported by agencies including the National Science Foundation, the Federal Aviation Administration, the Office of Naval Research, and the Air Force Office of Scientific Research. "V+ was up and running in a week," said CTC associate director for systems Dave Lifka. "Early runs are showing up to a 50% increase in performance over the Velocity1 cluster for some applications. This is exciting because such a cost effective solution makes it possible for us to continue providing our users with leading-edge systems more quickly than ever."

V+ strategic applications will be selected through a special allocations process. CTC is a high-performance computing and interdisciplinary research center located at Cornell University. CTC receives funding from Cornell University, New York State, a number of federal agencies, and Corporate Program members. CTC's AC3 is a consortium dedicated to the effective planning, implementation, and performance of industry standard systems, software and tools.

For further information about CTC, please see: http://www.cac.cornell.edu/