Blue Gene: IBM's Dream Machine
DECEMBER 26, 2002
NEWS ANALYSIS:TECHNOLOGY
Blue Gene: IBM's Dream Machine
Now in the making, Big Blue's hottest supercomputer ever will advance the state of the art and tackle some thorny problems
Ever since computer pioneer Seymour Cray built the first ultrafast machines in 1964, supercomputers have been the industry's version of a Formula One race. Unlike any other sector of high tech, companies -- and nations -- vie for the bragging rights to who has the most powerful machine on earth. It's a matter of national pride, global economics, cutting-edge computer science, and, of course, deep pockets.
"We're not striving for mediocrity, we're striving to be a leader," says Nicholas Donofrio, IBM's senior vice-president for technology and manufacturing. In the race to build the swiftest computers, "there's an aura that says you're on the leading edge," he says.
"IMPRESSIVE AS HELL." Blue Gene is IBM's Formula One racer. Nearly three years ago, Big Blue declared that it was going to push the edge of computer science through a five-year, $100 million research effort to set a new speed record for turbocharged computing -- a petaflop, or a thousand-trillion floating-point calculations per second. Blue Gene would solve some of the thorniest problems in life sciences. And on the way to building that machine, IBM would tackle some of the toughest problems in building ultrafast computers.
To reach that goal, IBM must advance the state of art in computer design and software development for large-scale systems. After all, the machine, set to be completed by mid-decade, would have 1 million microprocessors working feverishly on the most complex problems scientists could throw at the computer. "It's as impressive as hell," says Debra Goldfarb, an analyst with researcher IDC Corp. "It shows IBM is looking at different ways to solve difficult problems. Very few companies have the infrastructure and the vision to do something like that."
The payoff for IBM will be applying what it learns from the Blue Gene project to the design and development of all its computers. Blue Gene is going to require almost entirely new technologies from creating new techniques for building chips to developing new software to making sure the machine runs without a glitch.
SELF-HEALERS. Indeed, Blue Gene is associated with another research project at IBM called autonomic computing. The idea, like the human autonomic system it derives its name from, is to design computer systems that are self-healing, self- managing, and self-configuring. Elements of those software technologies are just starting to work their way into IBM products, but the ultimate goal is to be able to manage global networks of computers more efficiently.
In a way, Blue Gene is so massive and complex that it's autonomic computing in a box. "The technology [used in Blue Gene] has to span a very wide gap," says IBM's Donofrio. "It's systems-level science, programming science, and power management. It has to run nonstop, without a hiccup."
Then IBM can take a victory lap.
By Ira Sager in New York
Copyright 2000-2002, by The McGraw-Hill Companies Inc. All rights reserved.
Terms of Use Privacy Policy
Used withe permission of businessweek.com
businessweek.com |
