IBM, Microsoft and Hewlett-Packard set aside their rivalry Tuesday to share the stage at the first International Conference on Autonomic Computing in New York, at which representatives from the three companies mapped out fairly similar and harmonious strategies for working toward self-managing IT systems.
IBM claims credit for coining the name "autonomic computing" three years ago, to describe its new focus on automation as the key to addressing the growing problem of computing infrastructure complexity. That initiative became a core element in the on-demand vision IBM unveiled in 2002, a campaign that prompted many other IT vendors to craft similar strategies. HP calls its spin the "adaptive enterprise" strategy, while Microsoft spoke Tuesday of its Dynamic Systems Initiative.
The vendors' hyperbole about where autonomic computing might lead is dazzling, and so are some of the projects researchers are exploring. Steve White, IBM Research's senior manager of autonomic computing, spoke of an IBM venture exploring the construction of systems from self-responsible components: A storage-system component, for example, would recognize that it requires a database and would go find and configure its own database component. Called "Unity," the project has a working prototype, White said.
The gains from such studies won't come quickly, though. IBM, HP and Microsoft all said they're introducing autonomic features gradually into their products and don't expect to deliver any all-in-one miracle technologies. Panel moderator Ken Birman, a Cornell University professor, quipped ruefully that discussions of autonomic computing call to mind an old techie joke: "The engineer is sitting at the side of the bed, talking to his girlfriend, talking about how great it's going to be."
Some gains from autonomic research have already crept into products. IBM's forthcoming DB2 update, code-named "Stinger," includes tools to tune the database to accommodate fluctuating workloads, while HP has workload and availability controllers available for use in virtual server environments.
Representatives from the three companies agreed about the need for more automatic computing technologies, and about the incremental nature of their expected progress toward that goal. The only topic to spark notably divergent views was the issue of standards. An audience member called for greater industry standardization in areas such as error-log generation formats, a position backed by IBM's representative but rebuffed by Microsoft senior architect Anders Vinberg.
"I think it's a hopeless request," Vinberg said. Instead of pushing for deeper standardization, customers should turn to data transformation and integration tools to get heterogeneous applications and data sources to interoperate, he said.
IBM's architect, Jeffrey Frey, retorted that tools are essential but so is vendor cooperation.
"I don't think it's futile," he said. "I think there has to be more willingness by the vendors who play here to adopt standards."
IBM had a leading role in coordinating this year's inaugural conference, which ran in parallel with the World Wide Web Conference, but Microsoft will take the lead in arranging next year's gathering, in Seattle.
The conference attracted about 150 participants, a turnout attendee Daniel Menasce deemed strong. Menasce, a computer science professor at George Mason University in Fairfax, Virginia, said he's pleased to see growing attention paid to the problem of complexity.
"The vendor who doesn't get involved in this area is going to regret it," he said. "I'm not sure about cooperation, but I think we're seeing synergy and a common understanding that they all need to address this."