X
Business

IBM Research: We're taking over

the IBM Research Division is following a new path aimed at building more intelligence into computing by creating more autonomous computers.
Written by John G. Spooner, Contributor
IBM is rethinking how it uses its research brainpower.

The company's IBM Research Division is following a new path aimed at building more intelligence into computing by creating more autonomous computers.

By building hardware and software that allows computers to act more independently, IBM (ibm) Research believes it can create systems that offer higher overall performance and cost less to operate, in part because they would be easier to administer.

"This is one of the big themes for us," said Paul Horn, senior vice president in charge of IBM Research. "Because it needs to be holistic, it needs to touch everything we do."

Horn will detail the new approach in a speech to the National Academy of Engineers at Harvard University on Thursday.

Among the efforts at IBM Research is one to design new types of processors. One type, essentially a server on a chip, would include multiple processor cores surrounded by large amounts of memory on a single piece of silicon. This chip could then be strung together with others in varying numbers, appropriate for anything from entry-level servers to supercomputers.

Possibly the most important element in making computers more autonomous is software. Software developed by IBM Research would add the smarts to these server-on-a-chip processors, allowing each individual chip to have the ability to better manage the computing tasks it was charged with or hand them off if necessary, Horn said.

IBM will incorporate the new approach in Blue Gene, the supercomputer it is building to assist in genomics and other biological research. The supercomputer will be capable of operating at petaflop speeds--a quadrillion calculations per second--and will rely on server-on-a-chip-style processors.

At the same time, IBM Research is working on a project dubbed Oceano that will create software with enough autonomy to manage most of its own functions. Oceano servers would, for example, manage their own workloads, automatically adjusting the amount of computing capacity dedicated to a certain task, without human intervention.

As a result of its new approach, IBM Research has also slightly reorganized its structure. The division, which employs about 3,000 people, has a vice president to work with each IBM product group. Horn augmented this structure by adding two interdisciplinary roles to the council of vice presidents.

IBM Research's Arvind Krishna, director of computing utilities and Internet infrastructure, and Bill Pulleyblank, director of exploratory server systems, have been charged with focusing research on the new autonomic initiative. They will, therefore, be responsible for ensuring autonomy finds its way into both the hardware and software used in new products.

Pulleyblank, who also manages IBM's Deep Computing Institute, runs the Blue Gene project and manages its team of researchers. Krishna stewards the Oceano project.

IBM Research is counting on the two projects to prove its autonomous computing strategy, giving IBM's products an edge that could help win future business.

"I think (autonomous computing) will determine where the future of IBM and the IT infrastructure goes," Horn said.

Editorial standards