The Impact of Network State AoI on Throughput in a Wireless SDN

Abstract

This work studies the role of Age of Information (AoI) in the network state updating process for wireless software defined networks (SDN). The SDN routers must routinely update their knowledge of the network state, which is used as a basis for making routing and scheduling decisions. However, the network updates require communication resources, so there is a tradeoff between the frequency of updates and maximum network throughput. We assume the network state is Markovian and no new observations are received in between updates, so the AoI of the network state information impacts the ability of the network to optimize its performance. We formulate the problem as a finite-horizon Partially Observable Markov Decision Process (POMDP) for each period. For a symmetric fading model of the network, we derive the limiting performance and an upper bound. To generate policies for a range of fixed time horizons, we use Monte Carlo planning-based POMDP solvers. Simulation of these policies show that there is a finite optimal update period that maximizes network throughput. In addition, we study non-uniform update intervals, which can yield even higher throughput if the interval is chosen based on the state observed. We conclude that AoI itself is not sufficient to characterize performance, but what matters is the AoI for the specific network state information.