Shawn P. McCarthy
By Shawn P. McCarthy
Do parallel processing by tapping into idle downtime of your PCs
Is your agency taking advantage of all the processing power in its massively parallel supercomputer?
You say your agency doesn't own a supercomputer? It sure does. Just look around the office. All the processors in the desktop machines at headquarters and at field offices across the country add up to a pretty powerful parallel-processing system. It's just a matter of getting the machines to work together.
Many people have pondered ways to harness the idle processing time squandered by basic office computers, but few have ever done much to round up the computing beast.
Luckily, now there's a model. It's as simple as a networked screen saver. The catch is, you don't get to see the results immediately. The power is there but not the speed.
The fascinating model for free, Internet-based parallel processing comes from the SETI@home project.''
The Search for Extra-Terrestrial Intelligence Institute in Mountain View, Calif., supports a computer-intensive project using the Arecibo radio telescope in Puerto Rico to listen for signals in space that might indicate alien life.
The third kind
SETI became famous as a subject of Carl Sagan's book Contact, which was made into a movie starring Jodie Foster. Unlike in the book, the real-life SETI has not found evidence of extraterrestrial beings. But scientists are still looking, even though funding has slowed to a trickle.
That's where the SETI@home project comes in. Coordinated by the University of California at Berkeley, not by SETI, SETI@home splits off chunks of the institute's collected data for analysis on volunteers' computers.''
During idle time, a volunteer's screen saver downloads a chunk of this data and works on it. It then uploads the results and downloads a new piece. What does the volunteer get out of the deal? A fairly cool screen saver and a chance to be the one who finally locates an alien signal. Colorful 3-D bar graphs bubble across the screen as the data is massaged. It's all automated, so volunteers don't have to do any setup for the data chunks.
The thousands of volunteers for a beta test and the recent launch downloaded the SETI@home screen saver from setiathome.ssl.berkeley.edu/. Enrollment in late May stood at 370,000. Imagine having a computer with 370,000 CPUs.
SETI data is perfect for this type of background computation, because there's no rush'the signals originated years ago.
SETI might as well split off chunks and give them to people, as it could never do all the processing itself. If this is a valid way to build a slow but vast supercomputer, what other data could be processed the same way?
What other activities could use hundreds of thousands of Internet-connected PCs as private incubators in idle moments?
Not weather data because it's too time-sensitive. But what about other data collected by satellites'waves, temperatures, animal migrations? Signal processing for voice recognition of foreign-language phone calls? Tax audits?
The model has been built, and it works. If your agency wants to take advantage of internal or external computers in the same way, read papers from SETI@home project director David P. Anderson at www.surf.com/~anderson/pubs.html.
But no code
SETI@home does not plan to release its source code, to prevent someone from finding a way to tamper with SETI results. But you can visit sites of some of the development partners to pick their brains about parallel computing schemes. Visit the University of California Digital Media Innovation program at www.dimi.ucsb.edu/.
Details on how the SETI@home data is gathered and dispersed for computation are available at setiathome.ssl.berkeley.edu/sciencepaper.html. Read the project plan at setiathome.ssl.berkeley.edu/#status. And view a 1997 paper proposing that ordinary PCs could crunch the vast SETI data sets at setiathome.ssl.berkeley.edu/woody_paper.html.''
''Shawn P. McCarthy designs search and navigation products for a Web search engine provider. E-mail him at firstname.lastname@example.org.