IBM is ramping up its recruitment efforts with application companies in an effort to make an emerging distributed computing technology, called grid computing, pervasive in corporations.
In 2005, the company expects to work with between 40 and 50 independent software vendors to make their applications run across a "grid" of interconnected servers and specialised software, said Kenneth King, vice-president of grid computing. Last year, about 25 or 30 software vendors used IBM testing centres to "grid enable" their applications.
King made the comments on Tuesday at a meeting on grid computing with application providers at IBM's testing centre in Waltham, Massachusetts. He said IBM has made nearly $1bn from grid-related projects, mostly from hardware, and the company expects revenue from grid computing to grow rapidly over the next five years.
Although use of the term varies, grid computing is the idea of several machines combining their processing power to tackle computing jobs. Many academic institutions already have computing grids in place.
IBM, along with several other companies, is trying to apply the notion of distributing computing workloads across many machines in the commercial world. Right now, some companies are starting to experiment with grids for specific applications, according to executives at IBM and some of its grid software partners.
Hewitt Associates, for example, was able to cut down on the amount of time it took to bring in a new customer to its business outsourcing services from three months to one month and reduce the time to run a program from six hours to six minutes, according to Herve Collin, vice-president of operations at Sefas Innovation, which supplies document creation software to Hewitt.
Sefas was able to grid-enable its application within about two months and now pitches the grid computing capabilities as a potential time and money saver to prospective clients.
"Our work in grid computing started with a very pinpoint problem: performance. But as we started looking at all the possibilities of grid, [we saw] that there's much more," Collin said. For example, grid computing software can be used to schedule jobs more efficiently, or as a method to back up one server.
One of the challenges to the adoption of grid computing is that there are immature software interoperability standards. That lack of standardisation means customers need to pay for custom integration work between different grid software products, said IBM's King.
Increasingly, grid-specific standards will merge with general distributed computing specifications, he said. In particular, WS Resource Framework, WS-Notification and WS-Addressing -- which are each being developed at different standards bodies -- will become more important over time.
Signs of maturity in standards reflect the growing interest in grid computing among customers and in the computing industry overall, King said.
"Application providers are starting to care and grid-enable their applications," King said. "That follows the same pattern as the Internet and Linux, so we know that grid technology and computing concepts will be serious business."
