Simulation and analysis of adaptive routing and flow control in wide area communication networks

Abstract

This thesis presents the development of new simulation and analytic models for the performance analysis of wide area communication networks. The models are used to analyse adaptive routing and flow control in fully connected circuit switched and sparsely connected packet switched networks. In particular the performance of routing algorithms derived from the L(_R-I) linear learning automata model are assessed for both types of network. A novel architecture using the INMOS Transputer is constructed for simulation of both circuit and packet switched networks in a loosely coupled multi- microprocessor environment. The network topology is mapped onto an identically configured array of processing centres to overcome the processing bottleneck of conventional Von Neumann architecture machines. Previous analytic work in circuit switched work is extended to include both asymmetrical networks and adaptive routing policies. In the analysis of packet switched networks analytic models of adaptive routing and flow control are integrated to produce a powerful, integrated environment for performance analysis The work concludes that routing algorithms based on linear learning automata have significant potential in both fully connected circuit switched networks and sparsely connected packet switched networks.