Distributed Packet Forwarding and Caching Based on Stochastic Network Utility Maximization

Cache-enabled network architecture has great potential for enhancing the efficiency of content distribution as well as reducing the network congestion. This, in turn, has called for joint optimization of traffic engineering and caching strategies while considering both network congestion and content demands. In this paper, we present a distributed framework for joint request/data forwarding and dynamic cache placement in cache-enabled networks. Specifically, to retrieve the information about content demands and network congestion over the network, we establish a dual queue system for both requests and data, and define a dynamic mapping between the two queues with the help of dummy data such that the nodes can determine packet forwarding and caching strategies based only on local information. As the local objective function associated with Lyapunov optimization is time-varying due to the stochastic evolution of request/data queues, we develop a low-complexity distributed forwarding and caching algorithm via stochastic network utility maximization. We also prove the proposed algorithm achieves queue stability, and derive its region stability property for time-varying local optimization to demonstrate the convergence behavior. The simulation results verify queue stability and shows the proposed algorithm outperforms the existing ones.