Lehigh’s scientific computing infrastructure expanded by almost 50 percent recently when JPMorgan Chase gave the university a computer cluster of 96 Intel processors with 288 total gigabytes of random access memory (RAM).
JPMorgan Chase, with headquarters in New York City, is a worldwide leader in investment banking, financial services, financial transaction processing, asset and wealth management, and private equity.
The new equipment is expected to prove particularly beneficial to students and faculty doing research in engineering, science and mathematics.
“This is an extremely timely donation,” says Bruce Taggart, vice provost for library and technology services (LTS). “It will enable us to expand our high-performance computing infrastructure and provide faculty, researchers, and graduate and undergraduate students the best computing resources.”
“Four years ago,” says Taggart, “we embarked on a major effort to improve our grid cluster for research and teaching. JPMorgan Chase has helped make it possible to move to the next level in this campaign.”
The cluster was manufactured by Egenera Inc., a Massachusetts-based company that specializes in providing high volume data-processing equipment to businesses and other large clients.
A practical approach to education
It all started when last summer, Norman C. Brandel, vice president and business manager of JPMorgan’s Foundational Components technology group, phoned Lehigh to talk about donating the computer cluster.
JPMorgan Chase, says Brandel, was looking for a school or research institution that had a need for high-performance computing equipment and understood how such a gift could be used.
Brandel says that, in a sense, the technology itself made the gift possible. JPMorgan has developed one of the most advanced grid computing systems in the world. Called the Compute Backbone, the system has more than 2,700 CPUs in production, which are used to crunch complex risk management calculations. The system is the largest commercial supercomputer on Wall Street and one of the largest anywhere. Rapid advances in technology mean that JPMorgan is constantly upgrading and updating its high-performance computing equipment.
“The question for us,” says Brandel, “was what could we do with some of our older equipment, which is newer and more powerful than just about anything currently available? It’s kind of like when someone with a Ferrari buys a new one. What do you do with, as they say now, the ‘pre-owned’ Ferrari?”
Brandel says JPMorgan Chase chose Lehigh because the university has well-recognized strengths in engineering, science and math, and because the university supports research requiring intensive computing capability.
“We wanted to give the processors to a school that could truly use them,” says Brandel. “We know that Lehigh takes a very hands-on, practical approach to education. And we have worked with Lehigh graduates and found them to be very capable.”
Enlightened self-interest also played a role in making the donation to Lehigh, Brandel says. The expanded computing capacity will help Lehigh produce more “grid engineers,” who are in constant demand by JP Morgan and similar companies.
“The students who use these machines might very well end up becoming grid engineers, which would benefit us.”
Gale Fritsche, the LTS team leader for scientific and desktop computing, says the new cluster will prove particularly valuable to those doing research in optimization, computational materials modeling, molecular modeling, gene modeling and fluid dynamics.
Researchers in these areas come from computer science, physics, chemistry, biology, engineering and mathematics.
The gift has a market value of about $350,000, Fritsche says.
“This new equipment is state-of-the-art. The capacity it provides will open up new possibilities for research and enable us to better balance application usage.”
The new equipment will also ease the pressure on Lehigh’s SGI 3800, a 32-processor system that now runs at 80-percent capacity 24 hours a day. Fritsche said a typical Unix system runs at only 18 to 20 percent.