An Internet-connected grid of potentially millions of personal computers is being launched by a group of UK-based universities, weather services and software companies in an effort to predict global climate.
As a collaboration of Oxford, the Open University, Met Office, the Rutherford Appleton Laboratory and other institutions, the project that began this week will deliver the most accurate weather forecast ever, researchers said.
Met Office climate scientist Mat Collins, who has been working on the project for nearly four years, said the Internet-based grid will allow researchers to look beyond climate models that previously have been limited by computing resources.
“This experiment will give us the most comprehensive assessment of future climate change,” Collins said. “At the Met Office’s Hadley Centre, we can only perform a small number of projections of climate change at any one time. [The project] will allow us to run many more of these experiments and give better estimates of uncertainties in climate change to policy makers.”
Wanted: PC Power
Similar to the four-year-old Search for Extra Terrestrial Intelligence (SETI@home) project — which claims more than 4.6 million users who donate their PC power for research — the climate grid already has more than 1,000 people signed up, according to Collins.
Computer users anywhere in the world can join the experiment by visiting the project’s Web site, ClimatePrediction.net, and downloading and running a specially tailored version of Met Office’s global climate model software. Tessella Support Services managing director Kevin Gell said the project’s success depends on the computing power of millions of PCs.
The program runs as a “background process” and does not affect other computing processes, but users can watch the climate change model unfold before the experiment is completed and fed back to researchers via the Internet.
Weathering the Grid
Rob Batchelder, industry analyst and president of Relevance, told TechNewsWorld that because of its complexity, weather traditionally has been a problem for supercomputers to solve.
However, Batchelder said, there are weather problems that are “highly fragmentable,” making them ideal for grid computing.
“Grids are not good for data mining,” he said. “Grids are good for taking data and running lots of different models against it. That’s why grids are popular in financial services because there’s not that much data, but many different what-if scenarios.”
The brainchild of Oxford’s Myles Allen, the climate project will take advantage of advances in computing technology and leverage the power of today’s personal computers, which match the supercomputers of yesterday.
“It’s remarkable that the same climate models that were designed only a few years ago to run on the world’s best supercomputers can now be run on any up-to-date personal computer, making this experiment possible,” said Rutherford Appleton Laboratory’s Jamie Kettleborough.
Batchelder said other, similar grid projects are likely to be launched as this technology finds a broader base of applications and use.
Grids Get Hot
Yankee Group senior analyst Dana Gardner told TechNewsWorld that with new products such as Oracle’s 10g, Sun’s N1 offerings and similar work from IBM, the grid concept is “definitely catching on.”
While Gardner differentiated between the SETI@home and climate grid experiments versus corporate deployment of grid architecture in datacenters, he said both are uses of “horizontally” arranged resources.
“What allows this is not faster processors, but the network between and among them,” Gardner said. “The network, in this case, is the Internet.”
Batchelder disagreed with Hewlett-Packard CEO Carly Fiorina, who said at OracleWorld this week that grids have been “overhyped.”