IBM mixes data into grid computing recipe

Extending the concept of grid computing, IBM Corp. officials revealed details of new software research that engineers refer to as a data grid.

Most of the discussion around grid computing has focused on the potential to create a Great Global Grid that systems can use in a utility fashion to leverage hardware resources, similar to a gigantic virtual supercomputer. IBM’s research is building on several existing technologies, and some under development, to integrate data and applications in a grid-like scenario.

The data grid differs from a computing grid, in which systems can take advantage of hardware resources such as idle CPU cycles, although the two will be part of the same infrastructure. Data grids focus more on sharing information across large-scale computing infrastructures created by computing grids, according to Laura Haas, a distinguished engineer at IBM’s data management group in Armonk, N.Y.

Haas noted that bolstering integration for grids could help control explosive data growth, repurpose data pieces, and hook together data sources.

“This is building on our federated capabilities going more and more into large-scale distributed computing,” Haas said, adding that those technologies include the DB2 database, Data Joiners, and the forthcoming Xperanto technology. All are part of IBM’s strategy to access various data types.

“We see the federated capabilities as helping us in terms of how to make our systems play nicely in a grid of data sources. In some sense that is what federated does already,” Haas said.

Ron Schmelzer, a senior analyst at ZapThink, a research firm in Waltham, Mass., said that data grids are the next step in grid evolution.

“If you have the computing power on the grid, there will also be a need for data on the grid,” Schmelzer said. “In the long-term vision, people will start building applications specifically for the grid.”

He added that bringing data into the mix raises new issues that are not as prevalent on the hardware side. These include data management, security for distributing information, availability of data, and how to make money from grids.

IBM is not the only company developing grid computing technologies. Sun Microsystems Inc., Microsoft Corp. and Hewlett-Packard Co. are also working on grids, as was Compaq prior to its merger with HP. Several smaller players, including Platform Computing and Globus, are also on board.

“There is a small gang working on data for the grid, but as grid computing moves forward, the usual suspects will address it,” Schmelzer said.