Dynamic Cache Placement, Node Association, and Power Allocation in Fog Aided Networks

Proactively caching content at the edge is a promising way of alleviating the traffic load in the core networks, especially for latency sensitive video services. The content popularity herein is an effective measure for making efficient cache placement decisions. Most state-of-the-art works design the caching strategies under the assumption that video popularity is time-invariant. In practice, however, the popularity of different videos is constantly evolving over time, which will influence the quality of service and providers' operating efficiency. Therefore, in this paper, by considering a specified model of evolutionary video popularity, we investigate dynamic content placement, node association, and power allocation strategy in fog aided networks to maximize the time-average utility, meanwhile guaranteeing the queue in each edge node to remain stable. Utilizing the framework of the Lyapunov optimization, the formulated problem is transformed into an instantaneous mixed integer nonlinear programming problem at each time slot, which is then solved by our proposed iterative algorithm which jointly utilizes the greedy method and Generalized Benders decomposition. The simulation results demonstrate the effectiveness of our proposed algorithm and the tradeoff between the utility and average queue length.