Distributed and intelligent routing algorithm
A Network's topology and its routing algorithm are the key factors in determining the network performance. Therefore, in this thesis a generic model for implementing logical interconnection topologies in the software domain has been proposed to investigate the performance of the logical topologies and their routing algorithms for packet switched synchronous networks. A number of topologies are investigated using this model and a simple priority rule is developed to utilise the usage of the asymmetric 2 x 2 optical node. Although, logical topologies are ideal for optical (or any other) networks because of their relatively simple routing algorithms, there is a requirement for much more flexible algorithms that can be applied to arbitrary network topologies. Antnet is a software agent based routing algorithm that is influenced by the unsophisticated and individual ant's emergent behaviour. In this work a modified antnet algorithm for packet switched networks has been proposed that offers improvement in the packet throughput and the average delay time. Link usage information known as "evaporation" has also been introduced as an additional feedback signal to the algorithm to prevent stagnation within the network for the first time in the literature for the best our knowledge. Results show that, with "evaporation" the average delay experienced by the data packets is reduced nearly 30% compared to the original antnet routing algorithm for all cases when non-uniform traffic model is employed. The multiple ant colonies concept is also introduced and applied to packet switched networks for the first time which has increased the packet throughput. However, no improvement in the average packet delay is observed in this case. Furthermore, for the first time extensive analysis on the effect of a confidence parameter is produced here. A novel scheme which provides a more realistic implementation of the algorithms and flexibility to the programmer for simulating communication networks is proposed and used to implement these algorithms.