Home > News > Legion: Venture Won't Be Cheap

Legion: Venture Won't Be Cheap

From The Virginian-Pilot
January 1, 1996

By James Schultz

D octors could examine patients and their diagnostic tests from anywhere on the globe.

Folks at home would have instant access to huge digital film libraries.

From their laboratories, scientists could monitor, in real time and in detail, the globe's environmental health from a network of Earth-orbiting satellites.

Calculations that now take days could take minutes.

If you think the breakthrough software known as Legion sounds too good to be true, hold the pessimism.

"This isn't pie-in-the-sky. I'm very excited," said Dennis Gannon, a professor of computer science at Indiana University. "Legion is the avant-garde of a vast movement. Industry will be just inches behind this research. Within a year after Legion is up and running, there will be commercial versions."

Legion's creator is University of Virginia associate professor of computer science Andrew Grimshaw. He says the program would essentially turn the Internet into a virtual super-supercomputer, allowing researchers, engineers, businesspeople and ordinary citizens to user their computers to harness the computational muscle of dozens, perhaps hundreds of other computers around the world.

"There's a huge amount of computer power connected to the Net, most of which is unused at any one given time. If we're successful at this, we'll be able to borrow from other people's machines for a burst of computer power," explained computer expert William Wulf, AT&T professor of engineering at U.Va. who is working closely with Grimshaw on Legion development.

"If legion succeeds, it may become as or more important than the World Wide Web. Trust me: It's going to be wonderful."

Insiders call software such as Legion "middleware," because it acts as an intermediary between a user and a given task. With middleware, and for the person at the keyboard, only results matter. Legion would eliminate the need to master of be concerned with complex commands and codes.

"We have all these computerized resources -- high-performance computers, specialized databases, digital libraries -- but no way to use all this stuff in a seamless fashion," Grimshaw said. "We lack software to make all this hardware usable for real people. Legion would allow people to work together."

By most counts, Legion is one of just four projects of its kind in the world. Two other efforts are under way in this country: one in California and Illinois. The third, in the Netherlands, remains in the planning stage.

Legion is not being piggybacked onto existing software. As an original program, it is being tailored to work with a bewildering array of the several million institutional, business and government computers that together comprise the Internet. A prototype version of Legion is already running at three of four National Science Foundation supercomputer centers.

If Grimshaw, Wulf, and their collaborators manage to produce workable code, fast progress could be made on an array of vexing science problems. Researchers could come up with better models of everything from the creation of the universe to advanced rockets and airplanes.

Consumers and businesses should benefit from Legion's ability to turbocharge the computing power of workstations and personal computers. Routine access to supercomputer-like computational power could lower manufacturing design and material costs, for instance, drastically reducing the time required to move products to market and, ultimately, substantially slashing prices.

The government is certainly intrigued.

The Department of Energy's Los Alamos National Laboratory in New Mexico is members and 10 graduate students. John Reynders, a team leader in the Los Alamos Advanced Computer Laboratory, refuses to specify just how much cash his facility is spending.

Other federal agencies, including those connected with national security, may also be involved.

In all, says U.Va.'s Wulf, the venture will cost several million dollars. It will take 18 months to two years to come up with what Grimshaw calls "robust" software.

There remain daunting technical hurdles.

Aside from their one-key-fits-all-locks approach to the Internet, Legion designers are working on a top-notch security system to thwart hackers. Program architects are also devising ways to schedule and bill for computer time.

"We're striving very hard to keep this simple," Wulf said. "You can manage the intellectual complexity by only worrying about a few things. Big, complex systems don't work right."

Reynders of Los Alamos is among those convinced that Legion will conquer new territory.

"They're doing it right," he said. "Grimshaw's team is focusing on a very good design. At this stage it's very impressive."

If Legion or any of its competitors are to succeed, though, the speed at which data is sent over the Internet -- known as bandwidth -- must be dramatically increased. Those improvements are coming, Grimshaw and others say.

Grimshaw says that he would like to have a "usable" Legion system in operation by March 1997, and a more refined version ready by the time two new National Science Foundation supercomputer centers open in October 1997. The final product, says Indiana University's Gannon, might spawn new generations of machines and programs so powerful that they begin to function as cells in a vast, virtual brain.

"We're talking about a level of complexity that we couldn't have imagined. It's enormous," Gannon says. "When all that software starts interacting on our behalf -- and starts adapting to become more efficient in solving our problems -- we'll see all sorts of interesting behaviors emerge. It becomes something almost biological.

"I'm not ruling anything out."

Original Article | Local Copy


More news about Andrew S Grimshaw


Return To List