by Photo: IBM

Paul Horn

Recently, IBM distributed 75,000 copies of a manifesto entitled Autonomic Computing: IBM’s Perspective on the State of Information Technology (and posted it online). We asked the author, Paul Horn, senior vice president, IBM Research, to explain why his essay is so important.

Popular Science What, exactly, is autonomic computing?

Paul Horn Our nervous systems are autonomic: They free up our frontal lobes so we don’t have to monitor our heartbeats, breathing, or digestion. Autonomic computing systems can run themselves and figure out new ways of handling the workloads we put on them.

PS And this is something that we not only want, but actually need?

PH Urgently. The way computing systems are expanding, there soon won’t be enough people to keep them running. A typical company has database servers, Web servers, a connection to the Internet, a firewall, and all sorts of complicated goop. Each has, say, 20 parameters — that is, 20 knobs to turn — and they’re all interrelated. Getting in there and twiddling the knobs is too time-consuming. What’s needed is a holistic system, so that as we twiddle, we begin to create a system that manages itself.

PS How do you get that kind of system?

PH We can write optimization algorithms so that as the system’s workload goes up, it will automatically allocate more storage or spray information-processing requests to a server on the other side of the world that’s not being used. We need what we call an “immune system” to combat viruses on the Internet. Once it detects a virus, an immune system could send a cure out over the Internet faster than the virus could spread. But ultimately, we can make life simpler for users by, paradoxically, making an autonomic system that’s more complex. For instance, we have a project here at IBM called Blue Gene: the world’s biggest supercomputer. It has millions of independent streams of work going on simultaneously. Now, what happens when one of those streams gets clogged up and gives a dumb answer? Think of your PC for a moment: When something goes wrong, you do alt-control-delete and all that garbage and then reboot. Everyone hates it. But now imagine your system has, like Blue Gene, multiple parallel streams of work going on. If one gets clogged up, it just automatically goes to another stream. If you’ve got millions of streams . . . well, you can have a thousand or so that aren’t working. Obviously, this isn’t something that’s going to happen tomorrow. But we’re making progress in all the key areas.