Novel Fair Share Flow scheduling algorithm for ATM network

Abstract

ATM networks have many paths linking each host pair to achieve high bisection bandwidth for arbitrary communication patterns. Completely exploiting the bisection bandwidth may require flows between the same source and destination pair to take dissimilar paths to avoid hot spots. Though, the prevailing routing protocols have little support for load-sensitive adaptive routing [1]. This work proposes Fair Share Flow scheduling architecture for ATM networks. Fair Share Flow scheduling allows each end host to regulate traffic from overloaded paths to under loaded ones deprived of central coordination. We use open flow implementation and simulations to show that Fair Share Flow scheduling can efficiently use the network's bisection bandwidth. It outperforms prior solutions based on random flow-level scheduling, and performs correspondingly to previous work that assigns flows to paths using a centralized scheduler but deprived of its scaling limitation. We are using competitive game theory[3] to show that Fair Share Flow scheduling algorithm is more constant [2]. It makes growth in every step and converges to Nash equilibrium in finite steps.