ISARITHMIC FLOW-CONTROL USING LEARNING AUTOMATA

The main objective of flow control in a store-and forward packet switched network is a good tradeoff between throughput and delay. The isarithmic method is an algorithm for network access level flow control [6], that allows packets enter the subnet only if a free "permit" exists at the source-node. A learning automaton is situated at each exit-node, attempting to make an optimal decision for the distribution of permits. We assume a network with Virtual Circuits (VC) and we analyze the performance of the "Isarithmic-Learning" algorithm. An Event Driven Simulator has been derived for the comparison of "Isarithmic-Learning" algorithm with "Isarithmic-Random" algorithm (random distribution of permits).