Network routing algorithms can use a variety of metrics in determining optimal routing assignments. These algorithms may use static or dynamic information and have access to the relative or the absolute state of the network. The Genetic Algorithm (GA) is a heuristic search technique that has been used successfully to solve optimization problems in a number of domains. The ability of the GA to perform multi-parameter optimization based upon specific constraints makes it potentially useful as a network routing algorithm. This thesis analyzes the effectiveness of the GA as a dynamic network routing algorithm, and measures its ability to minimize both delay and frequency of packet loops in a network. These findings are compared to the performance of three other routing algorithms: Distance Vector (DV), modified Link-state (LS), and a random choice (RN) algorithm. The results demonstrate that the GA performs as well as the DV algorithm and better than the LS and RN algorithms at minimizing average network delay. Furthermore, the GA is superior to the RN algorithm at controlling the frequency of packet loops in the system. These findings validate that the GA is extremely effective as a dynamic network routing algorithm.

Pizza and soda will be served for those attending!