Print
Discuss
Send to friend
RSS feedsAbout 500,000 users are donating their idle computer time to the search for aliens in what is claimed to be the largest distributed processing project ever. The Seti@home (Search for extra terrestrial intelligence) software runs as a screensaver on Windows and Macintosh clients, and as a low-priority task on Unix machines. Using the internet, Seti@home is able to reach computers that would otherwise have wasted idle time animating toasters or displaying supermodels.
"The progress has been fantastic," said Seti@home project scientist Dan Werthimer. "Over 500,000 people are participating and contributing about 1,000 years of computing time every day. This is by far the world's largest supercomputer."
Screening for aliens
The project analyses data from the world's biggest radio telescope at Arecibo in Puerto Rico, piggybacking the long-running Serendip IV Seti programme. Each night the telescope scans the sky for signals, dumping 35GB of data onto tape.
Arecibo does not have a high-bandwidth internet connection, so this tape is sent by post to the University of California in Berkeley, where the data is split into 0.25MB blocks and sent out over the internet. Every five days for about five minutes, the client computer automatically exchanges data with the server when the user connects to the internet. All machines receive the same size work unit, regardless of their specification, and machines are sent new work once their batch is processed. The program consumes 20MB of disk space.
Each work unit represents about 24 hours of screensaver time on a 233MHz computer, but computers have been taking an average of 39 hours and 26 minutes to complete a work unit. Windows 95 machines have averaged a time as slow as 64 hours and 34 minutes. Pentium/Windows machines represent the greatest portion of the client base.
Are we alone?
Participants come from 203 countries, including many developing nations.
The top five countries are the US, the UK, Canada, Germany and Japan.
The US has more participants than the other four combined. "Their motivations vary," said Werthimer. "Most people are happy to help out. Many want to know are we alone? Is anybody out there? Some are computer-savvy and think this is a cool project to run on their computer. Lots of children are interested in Seti."
The server specification is Solaris on Sun Sparc. "Sun Microsystems donated the computers and we have extensive experience with these machines," said Werthimer. "The Unix-like operating system is very powerful. Most scientists are using this platform." The project has received further funding from The Planetary Society, Fujifilm Computer Products, Informix and Star Trek producers Paramount Pictures. But to date, there has been no government support.
Seti@home works on the assumption that any deliberate alien signals would be sent in a narrow frequency range because it is easier for us to filter from background noise. To eliminate signals from earth, the program looks for a signal that rises and falls in the 12 seconds it takes the Arecibo dish to scan the sky. Each work unit contains 100 seconds of data and the units overlap to ensure that no signal patterns are broken.
Werthimer admits the screensaver's appearance could be a key factor in the project's success. "Many users probably like the graphics and wouldn't run it unless it looked cool," he conceded.
Getting the ET signal
Anyone finding the crucial blip on their screen will be named as a co-discoverer after extensive efforts to confirm the extraterrestrial origin of the signal. The signal would be verified by the Seti@home team using the Arecibo telescope, scanning the same sky area. If confirmed, the signals are followed-up with dedicated telescope time. Another Seti group would then be invited to confirm the results independently, using a different telescope ensuring the signal is interstellar. If it is still unexplained, Seti@home would announce the signal to the astronomical community, using the internet to publish information on the discovery.
A team of 10 volunteers at Berkeley wrote the software over two-and-a-half year period. The source code is being kept confidential. "We hope to launch Seti@home II," said Werthimer, "and for security reasons, and to make sure everybody does exactly the same computation, we can't give out the source code." The data must be consistent if it is to have scientific value and the project team is wary of Seti@home becoming a hoax target for hackers.
The computing resources available have enabled Seti@home to search for signals about 10 times weaker than Serendip IV, although the project is still only processing 2.5MHz out of Serendip's 100MHz bandwidth.
The project is set to run for two years. Werthimer knows the chance of finding an intelligent signal is remote, but has a reply ready just-in-case: "Please send all your music, literature, science and information about the other civilisations you've been talking to. And, please, send information on how to hook up to the galactic internet."
To download the client software and sign-up, visit the Seti@home website at http://setiathome.ssl.berkeley.edu or check out The Planetary Society website at http://planetary.org.
MORE DOWN-TO-EARTH DISTRIBUTED NETWORKS
Although the largest, Seti@home is not the first project to exploit the internet for distributed processing. In January, Distributed.net won a contest sponsored by RSA Labs to crack a message coded using Data Encryption Standard (DES). The effort took 24 hours and used a dedicated DES cracking machine developed by the Electronic Frontier Foundation and the idle time on 100,000 computers worldwide.
Rather than attacking the algorithm, the solution used brute force to try every possible key until the code was unlocked. Because of the number of machines involved, the project could test over 250 billion keys every second. If each key were printed out and stacked, the pile would have grown 1,530 miles each second. Distributed.net has now taken up RSA's latest challenge to crack a series of DES messages coded with keys from 40 to 128bits. Each message begins with a known introduction. Computer users can download the client software and start hacking the code at www.distributed.net.
Apart from helping wannabe FBI agents with their desktop alien-hunting and code cracking, distributed processing has also been widely used by mathematicians. The Great Internet Mersenne Prime Search (Gimps) is using the internet to co-ordinate a search for the largest known prime number using an algorithm invented by seventeenth century friar Marin Mersenne.
The current record has over 895,000 digits. Although the search is inspired mainly by academic interest, prime numbers form the basis of PGP encryption and primes of this magnitude could eventually have computing applications.
The software and links to a range of other projects can be found at www.mersenne.org.
Another corner of prime number research has been explored using the internet by an international team comprising Harvey Dubner (US), Tony Forbes (UK), Nik Lygeros (France), Michel Mizony (France) and Paul Zimmermann (France).
They developed Windows software to find a series of 10 prime numbers an equal distance apart.
The result was discovered by Austrian mathematician Manfred Toplic using the client software after 'only' about 48 trillion numbers were tested, according to Tony Forbes' website at www.ltkz.demon.co.uk/ar2/10primes.htm.
The starting prime has 92 digits and the chain is made of primes 210 digits apart.
Colin Percival has turned his attention to pi, creating a program called PiHex to calculate the quadrillionth bit (250-trillionth hexit) of pi.
The project uses an equation that does not require all digits up to that point to be calculated. The program output the 40-trillionth bit to be zero in February this year. This work involved 126 computers from 18 different countries working from 19 April 1998 to 9 February 1999.
The calculation used 84,500 hours of idle time slices under Windows 95 and Windows NT. For more visit www.cecm.sfu.ca/projects/pihex/index.html.
See also:
All Network Infrastructure
del.icio.us
Digg this
reddit!
