A network calculus approach to throughput analysis of stochastic multi-channel networks

In this paper, two types of throughput, which are transient throughput and delay-constraint throughput, are investigated in a Gilbert-Elliott multi-channel network by using the stochastic network calculus. We propose a multi-channel model and derive an equivalent stochastic service curve guaranteed by the whole network. With the stochastic service curve, we obtain the lower bound of the network transient throughput. The throughput is non-asymptotic in that it holds for any number of channels and also fully accounts for transient regime. After, we derive the probabilistic delay bound of the multi-channel network, with which we further ascertain the delay-constraint throughput. Finally, numerical results are presented to show the impacts of the channel memory and number of channel on the transient throughput and that of the delay constraint on the maximum arrival rate sustained by the network.