Sun foresees global pool of computing power

At a Sun media event in New York, Chief Operating Officer Jonathan Schwartz unveiled a service by which customers can run some computing jobs on Sun machines for a cost of $1 per processor per hour.

So far, the technology is limited in scope, but Sun hopes that it eventually will blossom so that organizations such as stock exchanges with extra processing power can sell it back to a computing grid in the same manner that homes with solar panels can sell power back to the electrical power grid.

"If the exchanges are dormant at night, they can feed capacity back to the network," Schwartz said. He cautioned that such a vision requires security technology to protect computing tasks from tampering.

But such challenges will be solved, and Sun hopes to profit by running computers others can use.

"How big could the market be for this service? Add up the total number of hours used by all computer users on the planet," Schwartz said. "In the long run, all computing will be done this way."

Sun's not alone in having a vision of utility computing. Indeed, IBM has been working on grid protocols for years, Hewlett-Packard is building grid support into all its products, and both already operate complicated data centers for customers.

But Sun, punished by three years of market share losses and shrinking revenues, could claim some success if its plan just makes it relevant again in customer planning, some say.

"The notion is to make...customers and competitors stand back and say, 'What do you have to offer?'" said IDC analyst Vernon Turner. "Right now, this is a starting block for them. They're trying to get themselves back into the discussion."

Sun's initial service is geared for a limited set of high-performance computing jobs, such as detailing animated movie frames, analyzing investment portfolios or extracting oil field maps from seismic experiment data--all jobs that can run on an isolated set of computers. IBM and Hewlett-Packard already offer such services, though with different pricing plans.

Schwartz believes that general-purpose business computing tasks will come later, though the speed of light and other networking lags mean that geography hobbles many transaction-processing tasks that demand fast responses. "Over time, we'll look at the technology hurdles necessary to get to a true service grid," he said.

In this vision, Sun expects to run data centers packed with computers, but not generally to sell the computing power to customers. "We're looking at partners to deliver retail services to customers," he said.

Sun will rely on those partners for expertise. For example, Sun already has a partnership with SchlumbergerSema, which sells access to Sun equipment to customers in energy, finance, telecommunications and government customers.

In coming weeks, the company plans to offer more sophisticated--and probably more expensive--services that include partners with specific expertise. Partners will include CGI Group, Atos Origin and Electronic Data Systems, Sun said.