Enabling autonomic computing will require standardisation and
co-operation throughout the computer industry, said Robert Morris,
IBM vice-president of personal systems and storage.
Morris, who was speaking during a keynote presentation at the
O'Reilly Emerging Technology Conference in California yesterday,
described autonomic computing as a scenario in which systems are
self-managing, self-optimising, and do not require the amount of
expensive IT services needed today.
Complexity has created a situation in which it costs more to manage
systems than to purchase them, Morris said, citing storage as a
specific example.
Levels of computing such as Internet firewalls, load balancing
systems, and Web caching are making systems even more complicated,
providing motivation for a shift to autonomic computing, where
systems would manage and mend themselves, he said.
"No one vendor is going to make all the hardware, the systems, the
applications," said Morris, who is also director of the IBM Almaden
Research Center.
"It's going to be counterproductive for our industry if anybody
tries to control," the effort, he added.
After the presentation, Morris cited the Globus Project's Open Grid
Services Architecture as a standard technology for autonomic
computing. But an initial standards initiative for the industry has
yet to be formally launched, he said.
Morris said companies such as Hewlett-Packard and Microsoft have
been involved in meetings on the subject.
He also pointed out that in IT, systems such as RAID,
high-availability servers, and virus management have provided
examples of self-management, and that component technology needs to
become more self-managing and self-optimising as the industry
progresses.
In the database arena, IBM has a project called LEO (Learning
Optimiser), which takes statistics from the database and learns
from "mistakes" Morris said, adding that IBM planned to add this to
the next version of the DB2 database, in Version 8.
IBM is also working on a research project to boost RAID systems
called "collective intelligence storage bricks", which makes copies
of data to compensate for performance hot spots, Morris said.
Another technology, called hypervisor, which provides for
management and sharing of system resources.