Variable window flow control and ergodic discretized learning algorithms for adaptive routing in data networks

Abstract

Abstract In this paper we present a scheme that combines two new adaptive algorithms for flow control and routing. The flow control algorithm increments or decrements the window size based on the value of the sessions' delay and the routing algorithm uses a new fast and accurate Learning Automaton presented in [23] to choose the minimum delay routes in the network. We compare the performance of this new scheme with similar ones discussed in the literature. The new scheme, being a delay-oriented one, has a lower delay, at about the same throughput than some schemes given in the literature, thus resulting in a higher power ( = throughput/delay).