UB Team's Model for Protecting High-Speed Networks Achieves High-Bandwidth Efficiency and Fast Recovery

ANAHEIM, Calif. -- When a link or node in a high-speed computer network goes down as the result of a software bug, hardware failure, an error by construction workers or even a natural disaster, mission-critical operations -- such as telemedicine and financial or national security transactions on the Internet -- can be severely jeopardized in seconds.

Companies that provide network services protect their institutional clients, such as large corporations and government labs, from these consequences by establishing backup connections in their networks.

But many existing schemes require significant amounts of bandwidth, leaving less available for carrying out the critical operations they are designed to support. Other systems that chew up less bandwidth require longer time periods to restore service, resulting in serious delays to the customer.

Researchers in the School of Engineering and Applied Sciences and the College of Arts and Sciences at the University at Buffalo have developed the first approach that achieves both high-bandwidth efficiency and fast recovery speeds in protecting mission-critical connections against high-speed network failures.

Chunming Qiao, Ph.D., associate professor of computer science and engineering at UB, presented the research here today (March 21, 2002) at Optical Fiber Communications (OFC) 2002. Qiao conducted the work with UB Visiting Scholar Yizhi Xiong, Ph.D., and UB doctoral candidate Dahai Xu.

"It's a balancing act," said Qiao of the new method that manages to achieve high-bandwidth efficiency and fast recovery speeds by introducing two new tunable parameters into their approach.

"By picking the right values of these two parameters, we came up with a better solution," he said.

The UB researchers' approach to the problem uses a well-known mathematical formulation of an optimization problem, called Integer Linear Programming.

"Using one parameter in our model optimizes bandwidth, which does tend to increase the time it takes to restore service, but the other parameter in our model more than compensates by selecting a much shorter backup path," Qiao explained.

The UB team used computer simulations of the system to demonstrate the concept. Qiao said the next step is to develop a software package based on the approach.