Putting lazy computers to work

Toby Howard

Our PCs are surprisingly lazy things. If you run a system monitor program, like Microsoft's WinTop, and carry on with your usual work (or play), it's amazing to see how little processing power many common tasks actually take. Stop for a breather, and suddenly your machine may be 97% idle. What if you could share, or even sell, your PC's spare capacity? If you have an Internet connection, you already can.

The idea of using the Internet to harness unused PC power has become known as "distributed computing", and first became news in 1996 with the GIMPS project, which searched for the largest prime numbers. It was quickly followed by the incredible success story of SETI@home, which analyses radio-telescope data for signals from extraterrestrial life.

SETI@home works on data from the the Arecibo telescope in Puerto Rico, which daily sends 50 Gb of data to the SETI@home lab in Berkeley, California. The raw data is chopped into manageable chunks, and distributed to screensavers running on cooperating PCs around the world. Nearly 60 million chunks have been farmed out for checking, but no extraterrestrial signal has been found -- yet. The project has the support of 2 million people in 226 countries, and has clocked up an amazing 285,000 years of computer time.

In the UK, Myles Allen of the Rutherford Appleton Laboratory is organising the "Casino-21" project, which will use distributed computing for simulating changes in the earth's climate. The project isn't yet operational, but you can register your interest now. Beware that this isn't a simulation for the faint-hearted: "It calls for you to install a unique, state-of-the art climate model on your home PC and keep it running, possibly for a year or more", says Allen.

Until recently, distributed computing has been a strictly non-commercial activity. But now people are beginning to see how to exploit the idea to make money. If there's a huge pool of unused processor power out there, and if companies can exploit that power to solve their problems, there's a natural role for a broker.

One such go-between is Popular Power, which intends to handle both non-profit and commercial applications. Its first project is to simulate the response of human antibodies to influenza vaccines. With subsequent commercial projects, Popular Power intends to pay those who contribute their spare PC cycles.

Parabon Computation of Firfax, Virginia, has just announced a pilot scheme in collaboration with the US National Cancer Institute. They aim to use distributed computing to analyse the complex data from experiments with anti-cancer drugs and their effects on gene expression in cells. Currently, such research is hindered by the enormous computing times -- sometimes months -- required to analyse the data. With a distributed approach, Parabon hopes to reduce the work to a matter of days.

There are limitations to what distributed computing can achieve. In particular, it only works if a problem can be solved by breaking it into a lot of simpler sub-problems, each of which can be solved separately. Problems that require a step-by-step approach to solution just aren't amenable.

And there are security issues: how can a company be sure its computation on your PC isn't being compromised by a rival company? And how can you be sure that your climate modelling screensaver isn't doing something unpleasant to your PC while you're asleep?

There will be solutions to these problems, and the prospects of distributed computing are exciting. Our lazy PCs may even end up doing something useful.

Toby Howard teaches at the University of Manchester.