Two Time-Scale Edge Caching and BS Association for Power-Delay Tradeoff in Multi-Cell Networks

More network operators have recently provided CDN (Content Delivery Network) services where traffic engineering techniques such as the BS association are jointly employed with content delivery. This deployment attempts to reduce the network operating cost and enhance the QoS (Quality of Service) of end users. Toward this end, we study the base station (BS) association and file caching problem considering the spatial diversity of the file popularity and the realistic timescale separation between the file caching and the BS association decisions in this paper. Our design aims to minimize the file delivery latency and operating power consumption in the cellular networks where the tradeoff between these conflicting objectives are controlled by a single parameter. The short time-scale BS association problem is solved by using the convex optimization technique for a given file caching solution. However, the long time-scale file caching problem considering the varying BS association decisions taken at the short time-scale is difficult to tackle. To solve this file caching problem, we prove and leverage the submodularity property of the underlying objective function to develop a greedy content caching algorithm which guarantees a constant approximation ratio of the optimal objective value. Via simulations using real-world datasets, we show that the proposed algorithms outperform file caching and BS association algorithms that do not consider the spatial diversity of the file popularity in terms of the power consumption and delay performance in the geographically heterogeneous file popularity scenario.