Using a supercomputer in David Rittenhouse Laboratory's second floor late last spring, doctors from the Children's Hospital of Pennsylvania took MRI scans of a child's brain in just 17 seconds, a feat that normally requires 17 minutes with no fidgeting.

"Like a blurred photograph, an MRI scan is useless if the patient moves during the taking," says Penn's physics professor Robert Hollebeek, principal investigator in charge of DRL's second floor supercomputing facility.

What made this drastic cut in time possible was Hollebeek's fast supercomputer receiving, processing, and sending back MRI data 100 times faster than the original computer of the MRI.

But for all its amazing quickness, Hollebeek's supercomputer has always faced a major obstacle. The amount of data is so big that the cost is prohibitive. "Most of the data we process usually originates somewhere else in Philadelphia or in the USA," he says. "Getting the data to us via phone lines costs a lot of money. Sending back the processed information in the form of pictures or maps adds to those costs. Receiving and sending astronomic amounts of data has been a bottleneck."

Now all that is about to change. The National Science Foundation has invited Hollebeek and his physics computing group, and a dozen groups from other U.S. universities, to test drive the next super-fast generation of the information highway.

In return, Hollebeek's group is competing with the others to figure out how to connect data intensive sites like his into a backbone of a network already in place that until now has consisted of only four supercomputing centers in New York, Colorado, California and Illinois. This backbone of a network can send and receive information about five times faster than the commercial Internet, and the research connection to the backbone will run 30 times faster than the previous connection. But the backbone must grow now into a larger network. Penn's group will be one of the first sites to join.

NSF expects the resulting network to produce breakthroughs in multimedia communications, in remote access to very large data collections, and in cooperative experiments in science.

"Data mining with these big, super-fast computers is a hot topic in business, medicine and research, because data mining means creating new knowledge from vast quantities of information, just like searching for tiny bits of gold in a stream bed," says Hollebeek. Penn's group won the SUPERCOMPUTING '95 Data Mining Award.

The users selected to join the new network, like Penn's group, use and process large amounts of data. Hollebeek's supercomputer originally was built to exchange instant results of physics experiments between Penn, the University of Illinois in Chicago, and the University of Maryland. By allowing Penn to be one of the newer sites in this super-super information highway, Hollebeek's group will be able to look for new uses and features of this super-fast technology.

In a separate development, Hollebeek's group also received a grant to double the size of its current supercomputer hardware, one from IBM, and one from the National Scalable Cluster Project, which will work with research groups from Penn in high speed networks, linguistics, robotics, cognitive sciences, medicine, human genome data bases, chemistry, astronomy, and physics, as well as with other researchers from the University of Illinois, the University of Maryland, and Drexel University,

Return to Compass Features for September 17, 1996