Although the adoption of grid computing is growing in enterprises, users may face obstacles finding software that can be easily licensed to run on a grid.
In many cases, software license models are ill adapted to run in grid environments that can quickly scale up or down depending on demand, say users. If a software license is based on CPU usage, for example, costs can quickly escalate as more processors are called into service.
Software vendors "cannot license their software around a true on-demand compute model," said Chris Bennett, group leader at Acxiom. The data integration company is using grid computing to manage resources across 4,000 mostly commodity-based servers and to gain processing speed. Acxiom processes some 45 billion records a month.
"Can we get out from under software vendors' powers? Can we make a truly scalable infrastructure where we are not at the whim of a software vendor?" Bennett asked at the Grid Today 2004 conference held in Philadelphia this week.
Bennett said that most of the software Acxiom is using on the grid includes either applications the company has built itself or open-source software. The company has about 300 software developers.
While grid computing is gradually expanding into the broader enterprise, it's still mostly used for technical, compute-intensive applications, attendees at the conference said. And since niche vendors often write applications for these systems, licensing issues often arise.
"Because the vendors that we work with are small, it's very difficult to convince them that they should change their licensing practices," said Jeffrey Mathers, director of the research and innovation group at Johnson & Johnson Pharmaceutical Research & Development.
But John Hurley, the director of grid evaluation and implementation at The Boeing Co., said he believes vendors are starting to address licensing issues. Grid "is new for them also," he said.
"One of the biggest problems of the grid is accounting," said Hurley. "How do you pay for things? How do you set charges?"
Users say grid computing is worth whatever licensing problems companies may face. Mathers noted that when Johnson & Johnson was studying molecules used in drug development in computer simulations on a 32-way server, the processing time took about three months. After moving those simulations to a grid that uses several hundred computing resources, processing time was cut to about two weeks.
Grid computing appears to be heading toward wider enterprise adoption. Steve Yatko, IT head of global R&D at Credit Suisse First Boston in New York, sees grid use as key to developing service-oriented computing across the enterprise that focuses on delivering services to business users. In that case, the technology becomes secondary, he said.
The value of IT "will come from managing the whole and not the individual components," said Yatko, whose IT department manages 20,000 desktops and 9,000 servers.
While some IT departments are trying to reduce the number of vendors they deal with and consolidate on a single technology platform, Yatko said he believes in using best-of-breed systems built to open standards that allow interoperability. That's necessary to "attack the complexity problem," he said, adding that vendor "partnering will be critical, more critical then ever before."
But before grid computing can get widely established, IT managers first need to "find ways to automate those areas that are very labor-intensive -- that's going to be the key for being able to afford these new technologies," Yatko said.
Robert Cohen, an economist at the Economic Strategy Institute in Washington, claimed that grid use will have a significant impact on U.S. companies in the coming years, boosting productivity by 25 percent in a number of industries, such as pharmaceutical and automotive, within six years.
Grid's "potential to change business processes and change efficiencies within companies is dramatic," said Cohen. "The companies that have begun to do it see it, and it's in their bottom line."