| Recent developments in telecommunications technology such as high capacity fiber-optic cables have provided economical benefits in terms of capacity concentration. As a result, modern telecommunication networks tend to have sparse topologies. A main concern in a high capacity, sparse network with few alternative paths between communication centers is catastrophic events, such as the total lost of links or nodes, which will have significant effects on performance and connectivity of the network. These concerns have increased interest in network survivability/reliability and motivate the research in this dissertation. Three aspects of network design problems are considered (i) survivability and performance, (ii) overall reliability, and (iii) performability. A multiobjective genetic algorithm approach was proposed to solve these problems for both capacitated and uncapacitated networks. The performance of two new simulation methods and a new bound to estimate overall network reliability and performability are compared with those from the literature. The performance of the proposed genetic algorithm methodology is compared to the results from the literature and to optimal solutions. It is shown that the approach developed in this dissertation is flexible, effective and workable for complex and large size network design problems. |
