Crash-proof computer tactic revealed by UK researchers

For a PC user, nothing chills the heart like the "Blue Screen of Death" in Windows. It means their computer has crashed.

Although crashes are as old as computers, some UK researchers may be taking the first steps toward sending blue screens to the same graveyard where 5.25-inch floppy disks are buried.

The boffins at University College in London (UCL) have made something they're calling a "systemic computer" that they say taps into the chaos found in nature to enable a computer to heal itself.

A chief cause of computer crashes, according to the researchers, is the way computers process the instructions in the programs they run. They do that sequentially, a step at a time. Disturb that sequence, and the computer jumps the track and crashes.

That's not how nature works. "Its processes are distributed, decentralized and probabilistic," a computer scientist working on the research, Peter Bentley, told New Scientist.

Nature, he added, is also fault tolerant, which is why biological systems can heal themselves.

Fault tolerance, of course, is nothing new to computing. Servers have had it for years. And as far back as 2001, Bill Gates was calling on PC makers to build the technology into Windows XP boxes.

How it works

The crash-proof computer is put together differently than a current off-the-shelf byte box, too. It's made up of a number of systems. Each system has its own memory and contains context-sensitive data. Not only does each system contain data, but each contains the instructions on what to do with that data based on context.

Moreover, multiple copies of instructions are located throughout the many systems in the computer. That allows the computer to fall back to a copy of an instructions set should a running version get corrupted.

And because each system has its own memory, crashes caused because some code can't access a particular memory address can be averted.

In order to add randomness to their computer's operation, the researchers have replaced the program counter found in a typical PC with a pseudo random number generator. That allows the computer's system to execute their instructions in parallel and without one system taking precedence over the other.

While that sounds like it shouldn't work, the researches say it works quite well, and they'll be showing just how well it works in April, when are scheduled to demonstrate their hardware at an evolvable systems conference in Singapore.

Other efforts

The UCL researchers aren't alone in trying to design computers that operate as bio systems. A pair of researchers—one at the University of Manchester, the other at the University of Southampton—have been working for more than 18 months on a project to cobble together a million ARM processor to simulate to simulate the activity of neurons in the human brain.

Unfortunately, the crash-proof computer being developed by the researchers isn't something typical computer users will see any time. The clue is in the developers' assessment of the practical applications of the research: it could allow drones to reprogram themselves to cope with combat damage.

Subscribe to the Power Tips Newsletter

Comments