Abstract

Fixed Alternate Routing (FAR) is one of the most common strategies for routing in WDM mesh networks, which is aimed at reducing the amount of dynamic link state dissemination and computation efforts. For implementing the FAR, basically, each network node is equipped with a routing table, in which the alternate paths to every other node in the network are pre-determined at the network planning stage. As a connection request arrives,one of the end nodes behaves as a coordinator, which looks up the routing table for the alternate paths, and sends probing packets along each of the alternate paths, for gathering dynamic link state and wavelength availability along the alternate paths. Routing and wavelength assignment (RWA) is performed according to the gathered link state information. Although strategies for performing RWA with alternate paths have been extensively studied, ^1,3,4 there is little effort put on how to pre-determine the alternate paths in each S-D pair.Among all the document we have surveyed, only¹ and³ has specified the derivation of alternate paths: a node-disjoint path pair and a complete set of k-shortest paths are defined for each S-D pair in the networks.An extensive simulation has been conducted in¹ and showed that the experiment using the k-shortest path set can improve the performance over the case with a node-disjoint path pair as the alternate paths. However, the study in¹ did not consider the efforts for maintaining each alternate path,and the cost (i.e.,computation latency, extra control packets, and computation resources at the coordinator nodes, etc.) of dealing with hundreds or thousands of alternate paths when a connection request arrives, due to the fact that a large number of alternate paths may yield a large amount of probing traffic and computation efforts. Since the use of k-shortest paths does not consider the resource overlapping between each paths, the marginal benefits of adopting a large number of alternate paths can be quite low. Besides, if the distribution of all the alternate resources in the network cannot spread homogeneously, a traffic bottleneck can be generated easily.

© 2002 Optical Society of America