Grant to Improve Monitoring of Computers Should Increase Reliability

Since designing absolute, fault-free software is not possible, systems that accurately monitor the operation of concurrent computer systems and warn users of problems at hand may be the best solution, says Aravinda Prasad Sistla, professor of computer science at the University of Illinois at Chicago.

Sistla just received a three-year, $485,000 grant from the National Science Foundation to develop monitoring system techniques that make sure machines of the future, which will rely on multiple computer inputs, work the way they should all the time, every time.

Using mathematical modeling techniques, Sistla hopes to develop methods that improve monitoring accuracy and reduce “false alarms” that can slow down a computer operation. He will also develop models that can check for unobservable signs of system failure based on a computer system's observable work or output. Finally, he will develop monitors that collaborate with a computer system that allows it to evolve and operate correctly.

Sistla wants to develop monitoring techniques that work both actively and passively.

“We're modeling the underlying system probabilistically, and then we're developing various accuracy measures,” he said. “These are metrics not previously thought of.”

The proposed research will build on Sistla's ongoing work in the field of distributed and concurrent computer systems to find ways of making sure such software works properly.

“When you install software in, say, a car or airplane, you assume the components supplying inputs to the software are working correctly. But in critical applications, such as driving or flying, you don't want to make assumptions. You want to make sure things are monitored and detected for any erroneous behavior,” he said.