IBM wants computers that help themselves

The company on Monday plans to announce that it has established a new Autonomic Computing group, which will act as Big Blue's hub for research and product development in autonomic computing. Autonomic computing is IBM's term for computers that incorporate abilities like self-diagnostics, which give them the capacity to operate on their own with little need of attention from a person.

"I expect (autonomic computing) to make really substantive changes...and free customers to think more about their business than their (computer) infrastructure," said Alan Ganek, the IBM vice president in charge of the group.

The Autonomic Computing group will work with IBM Research and IBM product groups for servers, software and PCs as well as IBM Global Services to create autonomic technology and then build and release products that incorporate it.

IBM has been working on autonomic computing for more than a year, and in that time it has incorporated the technology into several products such as its newest DB2 Version 8 database software and its P-Series servers. But the new Autonomic Computing group will be in charge of turning that effort up a few notches, spreading the technology further inside IBM's products and working with the rest of the computer industry. There, its job will be to develop open standards, creating an autonomic computing architecture that will allow autonomic computers and software to interact, regardless of their manufacturer.

Ultimately, autonomic computing is designed to make computer networks easier to manage and to improve their reliability through better design, IBM executives said. Without autonomic technologies, the cost of finding and hiring IT managers to keep networks running will escalate beyond control, researchers have said.

"The end game is to deliver a computing environment that is online all the time as a utility," said Nick Donofrio, senior vice president of technology and manufacturing at IBM.

Building this kind of network harkens back to the move from a telephone system operated by a person working a switchboard to one that automatically routes calls. Autonomic networks would be able to sense a company's needs for bandwidth or storage and distribute them on the fly. The computers making up this network would each have built-in abilities to recognize, isolate and recover from problems, with as little human intervention as possible.

"It sounds far-fetched right now, only because it's a lot of hard work," Donofrio said. But ultimately "for consumers it means...incredibly available data and a much richer (online) environment."

This network would improve online auction sites, for example, by allowing a Web site to anticipate and adjust to handle spikes in traffic to avoid interrupting transactions.

IBM is not the only company working on autonomic computing projects. Sun Microsystems, one of IBM's main competitors in large servers, is creating its own technology dubbed N1, which will automate tasks often performed by hand and share computing resources across networks, the company has said. Hewlett-Packard is also working on similar products. These sorts of projects help a company sell both technology and services.

"In creating a division it will help to facilitate the amount of commitment required throughout the whole organization," said Mike Gilpin, an analyst with Giga Information Group. "It requires, for example, all the different parts of IBM Software to participate."

Making networks work
Some of the work of establishing autonomic networks is going on now. These kinds of networks are sometimes referred to as grids.

Grids, which connect one or more computer networks into a larger network, are a piece that needs to be in place to support wider autonomic computing initiatives, Donofrio said.