QoE-Driven Optimization for Video Services in Cache-Enabled Software-Defined HetNets

Abstract

To cope with the exponential increment of traffic data, caching has been seen as a promising way to relieve backhaul bandwidth burden in the Heterogeneous Network (HetNet). The Software-Defined Network (SDN) architecture allows Macro Base Station (MBS) to have a global view of the whole network to better conduct the cache update strategy. In conventional caching strategies, power allocation and user association are neglected, which may result in insufficient power for a user and cache waste in a small-cell base station (SBS). Quality of Experience (QoE) cannot be improved without taking into account power allocation and user allocation. In this paper, we jointly optimize caching, power allocation and user association to maximize users' average QoE for video services. Under the constraints of caching capacity, limited power and single association, a mixed integer nonlinear programming (MINLP) problem is formulated. Power discretization has transformed this problem into a Discrete Monotone Optimization one, and we propose a Joint Caching-Power-and-Association (JCPA) algorithm to obtain the global optimal solution. Also, a random projection algorithm is proposed to compute the lower bound faster in the JCPA's iterations. Simulation results have demonstrated the effectiveness of the joint optimization strategy in this paper compared to popularity-based strategy and least recently used (LRU) strategy.