Home computers now are as powerful as supercomputers of a decade past. In fact, it’s like using a new Ferrari to drive four blocks to the grocery store–e-mail and Web browsing, but no real workouts.
Indeed, computers mostly do nothing at all, a monumental wasted resource. Research universities years ago figured ways to tie idle computers together, giving them little pieces of huge problems to crunch, such as modeling climate change or designing chemical catalysts.
Seeing the value, savvy corporate information technology managers are adopting the concept for business applications that design airplanes or assess financial risk. And various organizations are recruiting home computers to help search for new cancer treatments, study human proteins and search for signs of intelligence in outer space.
IBM was among the sponsors that in 2004 launched the World Community Grid that encourages people to donate unused time on their home and office machines to help fight disease and do other socially useful research.
People who volunteer their computers can use their machines as usual, but when they’re asleep or out of the house, their computer will work on problems sent to it by the grid. One project looked for chemicals that could block the advance of the AIDS virus.
In that case, tiny slices of the quadrillions of computations needed were sent to individual computers for calculation. As each computer completed its task and sent back the answers, it got another problem.
Any time the computer’s user wanted to send an e-mail or find a Web page, the machine halted its grid work to do what its user wanted. In this fashion, an estimated five years of computations were completed in just six months.
Launched in 2004, the world grid has nearly 245,000 participants with more than 470,000 devices online. Most donors are individuals, but organizations including the Chicagoland Chamber of Commerce and United Way of Metro Chicago also donate computer time. So far the equivalent of 68,000 years of computing time has been donated for humanitarian research by members from 195 countries.
An impressive effort, it’s still only a drop in the bucket of what’s possible considering there are 650 million personal computers worldwide that could be linked to the grid.
Sometimes called distributed computing, many believe the popularity of grid computing will soar. Insight Research Corp., a Boonton, N.J., technology research firm, estimates the worldwide spending on grid technology and service totaled $1.8 billion this year and that it will grow to nearly $25 billion by 2011.
Corporate executives noticed that “computers are a thousand times more powerful than they were 15 years ago and they wondered, `Why aren’t we producing more with them?'” said Gerry McCartney, chief information officer at Purdue University. “If you think about it for more than 90 seconds, it doesn’t make sense.”
Like many research universities, Purdue has embraced grid computing enthusiastically. Some 4,300 Purdue computers, ranging from machines used by students to do homework to advanced research computers, are linked to receive assignments when they’re not working for their primary users.
On average, Purdue computers operate about 45 percent of the time on their primary work, 45 percent of the time on work assigned from the grid and are idle about 10 percent of the time. Millions of computing hours have been used by academic researchers at Purdue and elsewhere to design potential chemical catalysts, process images from electron microscopes and work on other problems.
In the earliest days of the Internet, some people conceived of it being a huge grid with all the computers interconnected and working together, but it developed differently, said Ian Foster, director of the computation institute at the University of Chicago and Argonne National Laboratory.
“People use the Web to communicate with other people through e-mail or with remote data,” said Foster, a pioneer in developing grid computing. “The Web isn’t for computers to communicate with computers.”
Enabling such communications requires special protocols and middleware that Foster and colleagues began developing in the mid-90s so that physicists and other researchers could distribute data to computers around the world to process their problems.
Several hurdles had to be overcome to establish academic grids. The software had to bridge boundaries between institutions and disciplines, said Foster. Those boundaries were crossed primarily because researchers were eager to get access to the computing power grid offered, he said.
Similar boundaries exist today even when a corporation attempts to establish a grid that’s limited to its own organization, said Steve Tuecke, chief technology officer of Univa Corp., a Lisle-based company that markets grid computing to corporate users.
Large pharmaceutical companies see the value of grids in helping their researchers collaborate on designing new drugs, Tuecke said. But linking researchers at centers scattered at disparate locations “encounters a lot of the same problems we ran into with academic institutions,” he said.
“Many large pharmaceutical companies grew through mergers and acquisitions. You have different operating systems, different security.”
Univa’s employees start with the Globus software written for academic researchers, which is open source and freely available, much like the Linux operating system. Univa staff customize Globus for corporate users. Early adopting industries include pharmaceuticals, semiconductors and aerospace, Tuecke said.
Just as widespread popularity of the Internet marked a new era in computing, grid computing will be part of the next era, said Robert Rosenberg, president of Insight Research.
“Imagine if the Web itself did become a grid,” said Rosenberg. “What a mind-blower if you had all the computing cycles worldwide available to you. Think about what could be done with that!”
———-
jvan@tribune.com
