Room 208 in the Wall Building doesn’t look much different from several other computer labs around the Coastal campus. It has 10 average-looking workstations lined across two walls, with 30 separate computers networked together. Nothing about the set-up suggests its power or its importance to the entire university.

This system, however, constitutes a bona fide “supercomputer” which, its creators propose, will significantly broaden the scope of the university’s research capabilities and allow Coastal entry into new fields of science that require the sort of mind-blowing number crunching that is especially associated with bioscience and DNA research.

John Graham, assistant professor of computer science

What exactly makes this system a supercomputer? Its principal architect, John Graham, who has been working on it, with student assistance, since he joined the faculty two years ago as an assistant professor of computer science, likens it to a toll booth on a superhighway. “You can have one booth taking care of one car at a time, or you can have 20 booths all working at once,” he said. “It’s all about ‘throughput.’ ”

Most computer processing, like most thinking, is done sequentially, explains Graham, but the supercomputer is defined by its ability to perform parallel processing. “If you have a million things to do you can get one person to do them one at a time – sequentially – or you can get a million people to do them at once – as in parallel programming.” Million is the operative word here. Supercomputers are designed to solve big problems, processing massive amounts of data which must be sorted, compared, contrasted and crunched simultaneously and in multiple combinations.

The computing, or throughput, capacity of supercomputers varies. Most PC users are familiar with Google, a sophisticated Internet search engine that employs parallel processing to find information on the World Wide Web. At the extreme high end are supercomputers like IBM’s Blue Pacific, which has 5,800 processors and operates about 15,000 times faster than a personal computer. Coastal’s supercomputer, says Graham, has 16 processors and 1,000 megaflops. (A megaflop is one million floating point operations per second – the average PC has one-half megaflop). Like other university computer builders working on a limited budget, Graham built Coastal’s system using off-the-shelf components and Linux operating systems – software which is distributed via the Internet at no cost – making it easy and economical to upgrade.

The installation of the supercomputer is being carried out in large part by Graham’s able cadre of student assistants, who are getting invaluable experience in the process. They are Mike Wisener, brothers John and Marcus Wu, Jason Pinkey, Ryan Karetas, Ken Ward and Jeff Vales. Graham creates the designs and sets the goals, but these students have wired most of the hardware, researched the software models and tested the installation. “They’re a pretty independent bunch,” says Graham. “If they don’t know something, they go out and learn it themselves.” Several of the students have given presentations about their work at the prestigious annual National Conference on Undergraduate Research.

Jason Pinkey, John Graham, Ryan Karetas and Jeff Vales

While the supercomputer in Wall 208 may not compare with those at MIT, it can compete with systems at most research universities and gives Coastal the computational clout equal to such schools as the College of William and Mary, Virginia Tech and the University of Delaware, where Graham got his Ph.D. and did his dissertation on distributed real-time multi-agents systems. Graham has written three books and more than a dozen journal articles on high-performance computing. In addition to his academic career, he also worked for 15 years as a research engineer for NASA, Sun MicroSystems, Eastman Kodak, Texas Instruments and other organizations.

Graham has a competitive spirit – he is coach of the Myrtle Beach rugby team, which, he is quick to point out, is “currently undefeated” – and he has set a goal to get Coastal’s new system on the University of Tennessee’s annual list of the 500 top supercomputers in the world by next year.

So why does Coastal need so much speed and power? Steve Sheel, chairman of Coastal’s Department of Computer Science, says Coastal is moving toward what is known in computer parlance as “Grand Challenge Problems.” These are defined as research areas with potentially broad scientific impact which require advanced, high performance, parallel processing technology. Examples of grand challenge problems include virtual reality, nuclear modeling, climate prediction, and two research areas in which Coastal is staking claims for the first time: DNA research and artificial intelligence.

Next page >>