Grid lets customers tap idle resources.
IBM this week is unveiling software in its WebSphere product line that promises to push grid computing further into the mainstream of enterprise IT.
The latest version of WebSphere Application Server, WebSphere 5.0.2 Enterprise Edition, includes grid computing features that let customers tap unused or underutilized resources across their enterprise and use them in the way utilities employ a grid of sources that can deliver power to users on demand.
WebSphere Application Server Version 5.0.2, to be available July 25, costs $30,000 per processor, officials said.
Stefan van Overtveldt, director of technical strategy for WebSphere at IBM, in Somers, N.Y., said the new software in WebSphere acts like a traffic cop and automatically directs and monitors system workload according to applications running and then routes traffic to an appropriate server to handle that workload.
"After a customer has a certain number of servers, they tend to spread the workload in a uniform fashion," van Overtveldt said. "This introduces situations where you can load one server heavily and the others can be at 10 to 20 percent capacity." But the new software "allows you to do allocation based on usage; it leverages grid concepts," he said.
IBM Research and IBMs software teams developed the technology over the last two years. It is the latest effort the company has made to work grid computing into the enterprise. In January, IBM announced 10 new grid solutions for five key markets.
Van Overtveldt said future versions of WebSphere will extend the grid scenario to be able to automatically coordinate multiple clusters of servers instead of single ones.
"Scientific grids typically deal with a single application," van Overtveldt said. "We support the dynamic allocation of a single application across multiple servers, and we will be able to support the prioritization of applications one over another."
Because server capacity is often underutilized, IBM added features such as WebSphere Performance Advisor, which uses live data to analyze changes and recommend actions to improve application performance. Another addition, support for automatic backup clusters, taps backup clusters if the primary cluster fails.
"The incorporation of grid protocols in products like IBMs new application server is evidence of the growing momentum toward commonplace grids," said Ian Foster, of Chicago, co-leader of the Globus Project for Argonne National Laboratory and the University of Chicago.
"While scientific and engineering researchers have been at the forefront of using these tools, now major IT companies and their customers are adopting the same standards, which promises to accelerate the grid communitys progress toward interoperability," Foster said.