High tech's latest bright idea: Shared computing

By Michelle Kessler, USA TODAY

SAN FRANCISCO — By day, Richard Vissa's PC sends e-mail, lets him surf the Web and handles other ordinary tasks. By night, when Vissa goes home, it becomes a super-powered computing machine, testing potential drugs for their ability to defeat disease.  "It's saving us millions," says Vissa, a computer executive with drugmaker Bristol-Myers Squibb.

The company is one of many dabbling in a new technology that many in tech, including IBM, say is one of the most promising new technologies in years. Dubbed shared computing, the technology allows companies to harness the processing power in every computer — at all times — and to combine it to crunch big computing tasks that before required expensive supercomputers. Shared computing also lets companies, or researchers, shift computing power via a network to where it's needed. That cuts costs and increases computing productivity.

While it will probably take five to 10 years for shared computing to go mainstream, experts say it will change how computers are used and could reorder the tech industry.

"It is the next big thing," says IBM Vice President Irving Wladawsky-Berger.

Swiss drugmaker Novartis is linking 2,700 PCs with shared-computing software to search genes for flaws that cause diseases such as Alzheimer's.

The investment — about $400,000 — should save Novartis $2 million or more, says Manuel Peitsch, head of technology

.

The pilot project at Bristol-Myers includes several thousand PCs in its research division. As in companies everywhere, they used to sit idle many hours a day and at night. Now they spring to life as soon as their users don't need them. That's because shared-computing software, built by Ontario software maker Platform Computing, instructs the idle PCs to connect with other idle computers on Bristol-Myers' network.

The idle computers are told to get to work on a huge list of computing tasks the company wants done to help it develop drugs. For instance, the computers might run simulations of how a proposed drug would react to a certain protein.

When done, the results go to a central database. A new task pops up. The computers keep working until their users return and command them to do something else, like send an e-mail. If a PC is especially powerful, it might even be able to send the e-mail and still crunch data in the background.

Bristol-Myers doesn't know whether it will expand the project or what its ultimate cost savings will be. There's good reason to be cautious. Linking computers to one another creates a security risk because it's tough to control who gets on the network. What's more, the software is still too buggy for many applications, companies say. And it has limitations.

Shared computing is best suited for tasks that can be divided into many small pieces, such as graphics rendering. Some sequential tasks, such as tallying retail sales data by the hour, will likely run best on a single computer, researchers say.

But eventually, small businesses and even consumers may be able to tap the processing power of a supercomputer as tech companies build shared-power networks for public use. PC maker Gateway recently announced plans to rent out the muscle of 7,800 PCs that sit idle at night in its stores. Someday, "You will plug into the wall (to get computing power) the same way you do to get electrical power," says University of Pennsylvania professor Robert Hollebeek.

Tech companies are betting on that.

IBM's new CEO, Samuel Palmisano, recently said shared computing and related technologies will drive IBM's growth — and the company made a $10 billion investment to prove it. Last year, Microsoft plowed $1 million into a shared-computing research organization called the Globus Project. Sun Microsystems is even giving away some of its shared-computing software to drum up interest.

Dell Computer, the No. 1 PC maker, also sells shared-computing software and, like IBM, will help companies figure out how to use it.

"We're pretty pleased with the opportunity to compete with supercomputers and mainframes," says CEO Michael Dell.

Gateway's plan is to rent out the processing power of the PCs it uses in stores for customer PC training and sales demonstrations. Through a partnership with shared-computing company United Devices of Austin, a small business could, for example, rent enough processing power to manipulate a large photo database. The cost: about 30 cents an hour for the processing power of a home PC. The company would go to a secure Web site to drop off its assignment and pick up its results.

Saves money, boosts efficiency

Other companies and researchers have started using shared computing to:

• Stretch resources. Walnut Creek, Calif.-based GlobeXplorer has 400 terabytes of data. That's about 30 times the amount of information stored in the Library of Congress.

The 3-year-old company has 100 powerful computers to handle the data that it sells — mainly aerial photographs of the Earth. Until GlobeXplorer installed shared-computing software last year, it struggled to keep up with demand on busy days.

That's because GlobeXplorer had two separate computer systems, one for sending photos to customers and another for manipulating them. On busy days, such as global news events, customers would flood the "send" system with requests. The other computers often were less busy.

Now, all the computers are on a shared-power network and any of them can do either task. GlobeXplorer's computers can handle three times the load they used to, says President Robert Shanks. He estimates the company invested about $55,000 in the technology — and saved close to $1 million. "It's a very efficient tool. ... It's unbelievable what people are going to be able to do with this stuff."

Shared computing can be especially helpful to small companies. Berkeley, Calif.-based pharmaceutical company Plexxikon in April linked its 40 computer servers to squeeze more power out of them. The system has worked so well that Plexxikon is considering also linking the PCs of its 55 employees to a shared-computing network, says President Kathleen Sereda Glaub.

• Send power where it's needed. New York investment bank J.P. Morgan's computing muscle is scattered. Most divisions and some overseas offices have their own banks of computer servers — which other departments can't access. A trader in New York, for example, may not be able to access a financial program running on J.P. Morgan's London computers.

That will soon change, thanks to J.P. Morgan's shared-computing program. When it's finished this year, bankers will have instant access to specialty computer programs used worldwide. That will make them more efficient, the company says, and let workers access computing power housed more places. "Time-to-market is key to us," says Steve Neiman, J.P. Morgan's head of high-performance computing.

• Store data and exchange data. As with processing power, shared-computing software also boosts storage. That's because it automatically places files wherever there is room on a network — and files aren't relegated to one computer.

A coalition of hospitals in England and the USA is partnering with researchers to share mammogram photographs. Today, most hospitals keep mammograms in-house, making it difficult for doctors outside the hospital to see them.

The shared-computing networks, developed by Oxford University, the University of Pennsylvania and other research groups, will increase storage and make it easier for doctors to access mammograms and compare them with millions of others. That will help cancer diagnosis, says Oxford University researcher Mike Brady, who heads the European project.

Potential is huge, path is hard

Research labs are pushing even more ambitious projects that could someday find their way to the business world. Purdue University professor Alok Chaturvedi has built a shared-computing network out of supercomputers to train rescue personnel to handle terrorist attacks. The computers create a virtual city under attack. Officials then decide how to handle it. In a recent exercise, 70 officials had to contain an outbreak of a fatal, contagious disease. The simulation is so complex that a single supercomputer couldn't do it. Charturvedi's shared-computing system has worked so well that he plans to beef it up with 70 of his department's PCs.

The University of Houston, too, is building a shared-computing network that it hopes to link with Texas A&M University this year. The computers will run weather simulations to try figure out how to make Houston less smoggy.

Despite the promise of shared computing, it's mixed news for the beleaguered tech industry. The companies that figure out how to make the software stable and safe "will make a lot of money," says Ian Foster, a shared-computing pioneer and University of Chicago professor. But companies may not need as many new computers if they use their existing ones more efficiently. One possibility is that more computing power will beget more uses for it — thus more computer sales.

"The payback is very substantial," says Songnian Zhou, chief technology officer of Platform Computing. But, he warns, it will take awhile for it to mature. "There's a lot of excitement, and some are promising the moon."