Most personal computers spend more time idle than active, so researchers are harnessing their latent processing power with “distributed computing.” In this model, a computer connected to the Internet performs a task and then sends the result back to a central server for analysis (see “Five Patents to Watch: Collective Computing,” TR May 2001). It’s an extremely powerful way to, say, sort through vast amounts of information looking for signs of alien life. But distributed strategies can’t handle complex calculations, which require teamwork: each PC must crunch its own data, swap results with the others and repeat the process hundreds or thousands of times. All those computers talking at once slows the calculations-and the Internet-to a crawl.
A system from Ottawa, Canada-based Internet research consortium Canarie could make the interactions between computers up to 20 times faster, expanding the scope of distributed computing. Project director Bill St. Arnaud has dubbed the system a “wavelength disk drive” because the exchanged data is stored in wavelengths of light circling in a fiber-optic network. Computers on the network can perform calculations and “write” the data to an assigned wavelength. They then “read” other processors’ results from the light stream, repeating the process until the calculation is done.
This spring Canarie installed the software for the system on computers in its pan-Canadian network. St. Arnaud expects to test the first application this month: a model to predict the progression of forest fires. In a pinch, local firefighting teams can jump on Canarie’s network, punch in their scenario and get a quick answer on where the blaze is headed and where they should focus their resources. “Most small communities in Canada are unlikely to have a supercomputer sitting around in the anticipation of this kind of emergency,” says St. Arnaud.