A methodology for the design of multi-channel network architectures

Two emerging architectures are opening up new opportunities in distributed network architectures. The first one is the feasibility of optical fibers that can offer enormous bandwidth. The second one is the availability of real-time tunable transceivers. In this thesis design of multi-channel network architectures (for local metropolitan area networks) is considered. The large scale design of multi-channel network architecture addresses a number of problems: the design of physical and virtual topologies, and the assignment of traffic flows to channels i.e., optimal routing. In this thesis these problems are formally proposed and solved. Main focus is on the virtual topology design problem that maximizes the sum of total traffic flow and minimum channel flow. Using the optimization technique of genetic algorithms, the virtual topology design problem has been solved. Design of a distributed heuristic for virtual topology design problem is also discussed. A methodology for the design of multi-channel network architectures is devised. Finally observations with the genetic (optimization) algorithm are discussed and continuing efforts in this area are outlined. (Abstract shortened by UMI.).