A self-vaccinating Internet, microelectronic devices tiny enough to be injected into your bloodstream, atomic Velcro -- sound like science fiction? Try science fact. Researchers have made significant breakthroughs on all three in the past year, and envision them changing the complexion of your infrastructure in three, 10 or 20 years.
Consider the "self-healing" network being built by engineers at IBM's T.J. Watson Research Center in Yorktown Heights, N.Y. Modeled after the human immune system, this is the likely future of virus protection and intrusion detection. Vulnerable end-user nodes at every company connected to the Internet are inoculated with antivirus code. When the code detects viruslike activity, it alerts a central agency computer -- a mainframe-type system likened to a doctor by Arvind Krishna, director of IBM's Foundations of Internet Research. That computer analyzes the virus, identifies its unique traits or signature, and promptly notifies all PCs on the self-healing network to block that signature. From discovery to fix in minutes.
Krishna thinks this immune system could eventually operate by identifying what each program should do -- rather than identifying acts that should not be allowed. If software publishers included affidavits of authorized outputs, the immune system could block activities that didn't conform to declared outputs.
IBM's first-generation self-healing network is in beta trials with Symantec, which has added the first versions of IBM's inoculation code to Norton AntiVirus 7.0 Corporate Edition. The immune system's effectiveness hinges on mass subscription to a central "doctor" service, which would be provided by ISPs. At best, it'll be a few years before any immune system scales to Internet heights.
The self-healing network is part of IBM's $300 million, five-year project to develop technologies needed to support billions of users and trillions of devices on the Internet. The project includes improved versions of capacity on demand, content distribution and metered services.
As billions of users come online, the 'Net will need bigger pipes. That's one of the challenges researchers at Sandia National Laboratories are tackling through their work on MicroElectroMechanical Systems (MEMS), devices measuring only a few microns -- smaller than a dust mite. They're working in an area of MEMS known as LIGA, a German acronym that translates to lithographic electroplating molding, says Jill Hruby, deputy director of Sandia's Materials Engineering Sciences Center in Livermore, Calif. LIGA devices perform precise mechanical duties, such as holding optical fibers in place for accurate positioning.
Cooler still, these devices are morphing from the strictly mechanical to the electromechanical. If made from the right optical material, they won't just hold the fiber, they'll become integrated circuits, Hruby explains. "In the future, these tiny devices are not only electrons but also photons and can move in an optical world . . . like tiny electrical computers," she says.
Not only would operators be able to use existing capacity more precisely but could also augment capacity with intelligence. Network management agents could, for example, become physical computers traveling through the wires. Surgery on humans could be conduucted by a microbe injection. Or for the truly wired, a cell phone could be implanted in your mouth.
If MEMS are too heavy, why not make chips from molecules? Researchers on almost a dozen projects are making headway doing this very thing.
For example, at the University of Illinois in Champaign-Urbana, researchers figured out how to place the nearly invisible molecules in a prescribed pattern on silicon. They sprayed a silicon surface with a one-atom-thick hydrogen layer and used a scanning microscope to pluck off unwanted molecules. "We've made atomic 'Velcro.' The atoms stick where you want, so you end up with an array of molecules," says Joe Lyding, professor of electric and computing engineering at the school. While this method of building a molecular integrated circuits is too slow for mass production, it proves a molecular chip can be created, and eventually manufactured -- perhaps in 15 to 20 years.
Meanwhile, researchers at Harvard University, Hewlett-Packard and IBM, among others, are building molecular-based memory, switches and other simple computers. The goal is a chip sporting a trillion transistors and 100,000 times more power than today's fastest CPU. "The chip can think, at least in a rudimentary sense," Lyding says.
Now, put these thinking integrated circuits together in a system, then link molecular systems and you'll have a network that can learn, talk, react, think, perhaps even do the dishes (with the help of a robotic arm). Gee whiz!
This story, "The best gee-whiz technologies" was originally published by Network World.