The search for extraterrestrials is on

The search for extraterrestrials is on. And it's not a Texas-sized mainframe leading the chase, but more than two million PCs and a willing host of amateur alien hunters

What would you think was the biggest computing project in the world? Something to do with simulating nuclear blasts, perhaps, or forecasting climate change, using some vast system that occupies acres of floor space?

What would you think was the biggest computing project in the world? Something to do with simulating nuclear blasts, perhaps, or forecasting climate change, using some vast system that occupies acres of floor space?

Not at all. It's the search for aliens - more precisely, any radio signals they might be beaming out. And it's being carried out not on some monstrous system, but by about 2.3 million personal computers scattered around the world, as part of a project begun by the Search for Extraterrestrial Intelligence (Seti) group. It's called "Seti@home", and consists of a small computer program which runs as a screensaver - that is, when you aren't using your computer - and chugs through the complex data analysis of radio signals received from outer space, looking for a steady signal that doesn't come from Earth and which lies in a particular range of frequencies that aliens would be likely to use - at least, according to our best guesses about what they might do.

Now, PCs - even 2.3 million of them - might not sound as though they could possibly stack up against the huge machines such as Asci White, the computer unveiled by IBM in July which covers an area equal to three tennis courts and can carry out 12 trillion calculations per second - more than three times faster than the recorded speed of any other computer, and 1,000 times more powerful than Deep Blue, the supercomputer that defeated chess champion Garry Kasparov in 1997.

But Asci White is only one machine (which will be used to simulate nuclear blasts mathematically). The statistics for the "Seti@home" project show that earlier this month its members' computers carried out 14.26 trillion operations per second - and that since the project launched in May 1999, its users have done a grand total of 3.7 hundred million trillion (3.7 ²⁰) operations. To reach that, Asci White would have to run non-stop for about nine months - but by that time, the Seti@home project would have outrun it even further, as the massed computers now have more computing power in total than the IBM behemoth. The statistics show that as more people join, they are also bringing more powerful computers to bear, so that in the course of the project's life the average time taken to process a chunk of data has fallen from 18hr 35min to 14hr 46min, for those using it in the past day.

The Seti project was probably the fastest-growing Net phenomenon (at least, until the music file-sharing program Napster came along). Released last year on 16 May, within 10 days Seti@home had 350,000 users in 203 countries; in just one day it added another 20,000. It passed the one million mark in September 1999, and two million this summer.

What also makes the Seti@home project remarkable - besides its size - is the fact that all the processing is being done completely voluntarily. Certainly, Seti itself could never pay for something comparable with Asci White. But by tapping into our fascination with the search for alien intelligence, and offering a simple solution, it has got people to hand over processing time that would otherwise go to waste.

The Seti package is a small download (taking about five minutes over a normal phone line) which installs itself. It then begins analysing a small packet of data - recorded originally by the Arecibo radio telescope - and once it has checked it for any constant signal suggesting alien intelligence, it passes the results back online to Berkeley. Various precautions are taken to make sure that users cannot fiddle the results; any data analysis which suggests alien contact, would be redone independently. But if your machine finds extraterrestrial life then you will get the credit, Seti insists.

Now, other organisations are hoping to tap into this powerful market for "distributed computing" to solve other complex problems which work better when broken into small pieces - such as unravelling the structure of proteins, finding potential drug candidates, and even predicting climate change. What might be even nicer, if your computer is usually just turned off at night, is that you could have the option of being the first person on the planet to spot a signal coming from an alien civilisation - or you might earn a few pounds letting those spare processor cycles solve problems for commercial companies.

David Anderson, formerly a computer science professor at the University of California at Berkeley who organised the Seti@home project, now works as chief technology officer for United Devices, a commercial spin-off. Based in Austin, Texas, it is now recruiting commercial companies interested in using idle computer time available over the internet.

Seti@home was not the first project to use distributed computing over the internet: that was almost certainly distributed.net (www.distributed.net), formed in 1997 to crack encryption keys to coded messages. Such messages are encoded by multiplying together two very large prime numbers. With the message authors' agreement, distributed.net would issue volunteers around the Net with a version of the message and a computer program for their PC telling it what range of prime numbers to try. Of course, mathematicians have for years been setting their PCs to work trying to find record prime numbers, and using the internet to communicate their results, but that has not been a centrally-coordinated project with distributed effort.

Other distributed projects now underway are seeking out Fermat numbers (of the form 1+2 ²ⁿ; the question is, are there any primes beyond n=4?) and optimal Golumb rulers, which are sets of non-negative integers where no two distinct pairs of numbers in the set have the same difference. (A five-mark Golumb ruler is 0,1, 4, 9, 11; finding them becomes exponentially harder for sets of 24 or more, but there are applications in X-ray crystallography, radio astronomy and coding theory.)

And certainly if you can find the right project, internet users will happily lend a hand. In autumn last year, Myles Allen of the Rutherford Appleton Laboratory in Oxfordshire had the idea of using distributed computing to try to solve climate simulation. He posted a message on the Net noting that the work would try to introduce "fuzzy prediction" to reflect the variation of risks and probabilities in the forecast, rather than just one "best guess".

"We don't have the computing resources to do this any other way. So, if you're lucky enough to have a powerful PC on your desk or at home, we're asking you to do your bit so the right decisions get taken on climate change," he asked. The "Casino-21" project site got 15,000 replies in two weeks.

But it is the commercial side which is booming now. Parabon Computation, a 45-person company in Fairfax, Virginia, aims soon to start farming out screensaver work in biotechnology, financial and pharmacology research; people who sign up will get a payment for their machine's time per unit of work done. You won't be a millionaire, but Parabon says it might be worth a few pounds a month.

Another company, Applied MetaComputing, has government customers such as Nasa and the US Defense Department. And another company, TurboLinux, sells a product which can distribute work within a company for its idle computers. Among the customers for that are Procter & Gamble, which uses it for computeraided design, and Motorola which uses it for chip design.

So the demand is certainly out there. The problem would-be distributed processing companies face though is that they have first to persuade their potential clients that their data will be safe out on the Net: "People with serious computations are not likely to trust results coming from unreliable machines owned by total strangers," said Bob Metcalfe, formerly of Xerox Parc, and the man credited with inventing the Ethernet networking system.

But Parabon thinks it has got around that problem, by offering software much like the Seti@home product. It hopes to launch a project soon for the University of Maryland to analyse protein folding, an essential step between genes and biochemistry. It is also working with the US National Cancer Institute's molecular pharmacology lab on a project to simulate how cancer cells react to different drugs. At the same time distributed.net is working with the UK's Sanger Centre near Cambridge on mapping the human genome: its community of 60,000 participants, with 200,000 computers, are equivalent to more than 180,000 Pentium II 266-MHz computers working flat out around the clock.

Who knows? It may be that important future discoveries won't be made on huge machines like Asci White, but on one close to you - perhaps the one on your desk. As Richard Crandall, a scientist at Apple Computer, comments: "Machines have no business sleeping."

But for now, the project to beat them all is still Seti@home. "It's the world's largest supercomputer," said Dan Werthimer, chief scientist at Seti@home and director of the Berkeley Seti program. "It's made our search 10 times more sensitive, so we can find weak signals and pulse signals, things we couldn't look for because we didn't have enough computing power."

Seti@home: http://setiathome.ssl.berkeley.edu/ The Ral Climate page: http://www.climate-dynamics.rl.ac.uk Parabon Computing: http://www.parabon.com Distributed.net: http://www.distributed.net