Performance Evaluation of A Class of Adaptive Bifurcated Routing Algorithms

Abstract

With the cost of computation decreasing, packet-switched computer communication networks are becoming increasingly cost effective. In this paper a class of bifurcated routing algorithms is considered. One of the algorithms uses distributed control and routing decisions based on stochastic measures. The second one is a deterministic routing algorithm that uses localised deterministic routing decisions based on centralised measures. The third algorithm uses "Learning Automata" principle which is a promising technique because of its simplicity and ease of implementation. The three algorithms have been modelled and simulated to evaluate their performance under network conditions. A quantitative investigation of the three algorithms under various traffic conditions has been carried out. The overall average time delay, the average retransmission probability, and the average response time have been taken as a common basis of the comparative study. Regarding these measures, the learning automata has proved to be the best.