Harvesting Computer Cycles—for Better or Worse?
As we continue to close in on the one-PC-per-student ratio—or perhaps even exceed it if you consider public computer labs that dot the landscape of laptop computer campuses—the muted chorus of an old Pete Seeger song floats through my subconscious: "Where have all the cycles gone, long time passing?"
Taylor-esque time-use studies would be useful now. How many hours a day do we use our computational powerhouses? The scale (and implications) might run something like this: less than 1 hour per day (potentially insufficient return on the investment); 1-3 hours per day (appropriate use—user has a life beyond the keyboard); 4-7 hours per day (either in the information industry or socially isolated); greater than 7 hours (user needs immediate help). Yet even the wildest extreme leaves plenty of hours in a 24-hour day. The rest of the time the computer either sits inert, unpowered, in hibernation or sleep mode (if it's a laptop), or it idles with perhaps a screen saver keeping phosphor paintings from being etched on glass. What a waste.
Of course, if I had been the first to recognize this, I probably wouldn't be writing this column. Creative, benign, and some downright devious ideas have been developed to exploit unused computing power in homes, offices and, yes, college campuses. The most famous is the SETI project—the Search for Extraterrestrial Intelligence. Volunteers the world over have signed up to donate their unused computing cycles by downloading a small program that allows the SETI group to process complex signals received from radio telescopes in search of meaningful sequences that might reveal life in the stars.
Connecting thousands of computers using the Internet as the network establishes a loosely coupled peer-to-peer computing environment. Separately, even the most powerful of these machines pales in comparison to the supercomputing potential of the entire network of PCs. Together they rival and even exceed truly big-iron supercomputers for some classes of problems. Commercial use is just getting started. What is being sold is computing power for research firms that need computational muscle.
Distributed computing isn't as robust as a dedicated supercomputer, but then it d'esn't cost as much, either. It's also not an either/or proposition to the user. Windows multi-tasking operating systems were developed because we tend to find value in running more than one application at a time—a browser and a word processor, for example—or, simply multiple browser windows. Peer computing programs run in the background. They use the cycles that are flying by at a gigahertz, while the human processor stares at the screen and clumsily types letters on the keyboard, represented on the LCD panel and refreshed at anywhere from 60 to 120 times per second.
Who is using distributed network computing? Cancer research and AIDS research, to name two beneficiaries. Who is joining these computer Borgs—the super intelligence that absorbs the people it conquers and joins their brains to the multi-body "Borg" entity? A trend in voluntary "pay for cycles" has made it attractive to anyone who wants to earn a few extra bucks, especially when the contribution g'es toward a good cause. But d'es it? With the "pay-for-cycles" model is potential for a reverse Robin Hood effect: Buy from the poorer masses and sell to the rich corporations.
Where are the high-performance computer networks with an abundance of PCs sitting idly a great deal of the time? This isn't necessarily found money. It takes time and effort to install and maintain peer-computing applications, which could conflict with security software installed to protect PCs from rogue applications or viruses. In short, cycling could be more of a pain than it's worth—or, worth a second look. Even small colleges and community colleges have labs that sit idle much of the time. These could become new research or educational tools. If you were able to navigate what might be treacherous ethical waters, they might even represent some revenue.
Wherever there is opportunity to do good, there is the potential for other uses. Consider the number of computer users on large commercial ISP networks such as AOL (x million users), MSN (y million users), or Juno (z million users). Now there are some potential cycles to harvest! Are they doing this now? Do you know if they are or not? D'es your disk light flash with activity when you aren't doing anything? D'es your PC sometimes seem to slow down for no apparent reason while you're just typing?
Perhaps as an ominous harbinger, the ISP Juno indicated last winter that it was thinking about including distributed computing software in its dial-up network client upgrades. It d'es this now for those wishing to volunteer to participate in distributed computing projects. More troubling was the suggestion that this could be implemented at any time, with a simple modification of the service agreement. By breaking the shrink-wrap packaging you might inadvertently join a distributed computing network, with your machine calling up an ISP in the middle of the night to chunk away on problems for who knows who. And, of course, any resulting phone bill or network usage charge would be yours. So beware of calls in the night. It's 10:00 p.m. Do you know where your cycles are being used?