Cooperative Content Caching For Software-Defined Radio Access Networks

With the prosperity of Mobile Internet and the popularization of smart devices, the explosive growth of data traffic has brought great pressure to the RANs, especially to the capacity of radio channel and backhaul links. Most of studies for RANs are focused on the network architecture, resource allocation and Fronthaul etc. Although it is possible to increase the capacity of cellular networks, the growth speed of network capacity lags far behind the ever increasing data traffic demands that are mainly contributed by videos. Studies show that in a certain time frame, the users' requests for high popularity contents often exhibit a centralized preferences. Considering the reuse properties of this traffic load, content caching technique in RAN is emerged.A multilayer heterogeneous RAN with SDN characteristics model is considered in this paper, where the macro-cell base station(MBS) coordinates the content caching performed at the small-cell base stations(SBSs). If the requested contents exist in the local cache, they can be delivered directly through a wireless link. Otherwise the MBS with centralized control function will allocate resources from other SBSs to achieve cooperative caching. Due to the limited cache capacity and cooperation among SBSs, we have to consider what and where to cache effectively among SBSs to enhance network performance.Optimal content caching has been studied and some basic principles about how to reduce content provisioning cost significantly has been proposed. Inspired by two benchmark cases which are completely independently caching and fully cooperative caching, a heuristic cooperative caching algorithm between them is proposed. Where the cache space at each SBS is partitioned into two portions for storing the duplicated and unique contents respectively, then the content placement problem is transformed into determining the suitable partition factor. For some specific scenarios, analytic expression of the optimal separation parameter is obtained. Simulation results show that there are always optimal separation parameters to significantly reduce the content provisioning cost.