Enterprises are mainly interested in grid computing because it promises to ensure that each computer uses all of its processing power, say analysts. Currently, most enterprise servers spend a great deal of time doing nothing. But when computing power is balanced across an internal grid, companies up their utilization rates and cut costs.
Eventually, through a combination of grid computing and Web services, companies will be able to partner and share resources, or even earn money by renting out their servers' downtime to any business in need of a boost in processing power.
Analysts point out, however, that this "grid-locked" world won't become a reality for quite some time. The idea of sharing servers, either internally or across enterprises, raises vexing concerns when it comes to security, application development and company politics. They say businesses will need to address these concerns before they can seriously venture into grid computing.
"The idea of linking together a collection of machines in the commercial environment with a degree of security and reliability that commercial customers want is a long ways off," said Rich Evans, a vice president with the Stamford, Conn.-based analysis firm Meta Group. "I mean people are having enough trouble now with security on straightforward lockdown systems."
Tony Iams, a senior analyst with D.H. Brown Associates in Port Chester, New York, said that before a company can set up a grid and improve utilization internally, they must first look closely at the demands of their applications.
"In the high performance technical computing space there are a number of applications that are embarrassingly parallel and they seem to work well in grid environment," said Iams. "The challenge today is to identify new applications, some of which may not be in the technical space that would work in the grid environment."
Currently, said Iams, commercial applications that work well on a grid include highly parallel modeling and simulation applications which might be found in a company's engineering or product development departments. For now, he said, traditional business applications such as those used for inventory management and database processing are not conducive to grid computing.
After a company examines the computing needs of its applications, Iams said the next step is to use that knowledge to decide which tools and standards you'll use to set up the grid. The tools a company will need to decide on include message passing systems and scheduling and queuing tools, which send jobs around the grid for processing. He said the standards being used in grid deployments are beginning to mature and include the Open Grid Services Architecture (OGSA).
"The OGSA is an industry standard developed to determine how a grid should be built and operated," Iams said. "It's designed to be agnostic in terms of your applications' APIs."
In terms of developing new applications for the grid, Iams advised that companies make use of Web services standards like XML, SOAP and UDDI.
"Grid computing is converging with Web services," he said. "If you use Web services in theory you should be ready for grid services as well."
Throughout all of the planning and deployment phases of grid computing, it's important to keep everyone in the company in the loop, Iams said. When people realize that they are being asked to share equipment with other departments they often get protective of their resources. Iams said one way to overcome this problem is to offer financial or budgetary incentives for sharing resources.
While businesses have a long way to go before grid computing becomes the standard, don't expect a slowdown in the onslaught of grid marketing messages from the major vendors, warned Meta Group's Evans.
"Vendors have to find something to sell," Evans said. "I think they're all looking for a little bit of pizzazz to get people enthused."
