The Internet suffers from long-lasting routing failures. Recently, the main Internet backbone was disabled for 8 hours. Some users depend on the network for their operations (think of eBay, Schwab, the San Francisco Airport). Our research looks at basic ways to make the network more reliable. Our approach is protection switching.
Various algorithms have been proposed to choose normal and protection paths. These algorithms are heuristics that attempt to optimize some performance measure such as blocking probability, throughput, or delay. We have explored various combinations of these methods to see whether they make a difference and to attempt to derive guidelines on how to choose them.
A related idea is the adaptive reservation of bandwidth. Since it takes some time to reserve bandwidth, it may not be possible to do it when needed. The idea is then to reserve a little bit more bandwidth in advance, in anticipation of the next requests. We study the trade-off between speed and inefficiency of such a scheme.
Here is a paper with Eric Chi and Rajarshi Gupta that compares different algorithms for normal and protection paths. This other paper with Eric Chi and Michael Fu studies how to to pro-active resource reservations. (See also this more complete version of the paper.)