Hierarchical Edge Caching Strategy For Software-defined Networks

With the rapid development of smart devices and mobile data services,the global data traffic has experienced an explosive increase.Since the capacity of backhaul links is limited,the data explosion will cause network congestion inevitably.How to break such bottleneck of capacity limited backhaul connections becomes a challenge for developing 5G networks.Albeit many and variable,numerous studies found that most of the traffic loads on the Internet are contributed by the repeated views and downloads of the same group of popular contents.Inspired by this,the backhaul traffic loads and network congestion can be significantly reduced by caching these popular contents at the network edge.In this paper,we consider hierarchical edge caching in a software-defined network with limited backhaul links,in which the control function is decoupled from data plane and centralized in control base station(TBS),and traffic base stations(TBS)only provide data coverage for users.In order to achieve content caching and delivery efficiently,we design both storage allocation and content placement between the CBS and distributed TBSs jointly,and present an analytical framework of minimizing the average content provisioning cost of the system under limited sum storage capacity constraint.This is NP-hard in general.As a comparison,a baseline non-cooperative caching strategy is first introduced where TBS and CBS cache popular content independently.However there always exists caching overlap in non-cooperative caching strategy.To further improve the storage resource utilization,an efficient hierarchical cooperative edge caching strategy is proposed by leveraging the vertical cooperation between TBSs and CBS,and horizontal cooperation between TBSs.Simulation results show that the average content provisioning cost of network can be significantly reduced by adopting our analytical optimal storage allocation factors between TBSs and CBS,and the proposed hierarchical cooperative edge caching strategy can achieve a better performance than non-cooperative caching strategy.Besides,an extra gain can be achieved by switching between the vertical and horizontal cooperative caching modes.In the above analysis,caching decision is only based on content popularity and ignoring the effect of content size.To further optimize the caching decision,this paper propose a size-weighted popularity(SWP)-based caching framework,where both content popularity and content size are taken into account when making caching decision.Depending on the available priori knowledge and caching time,a proactive caching algorithm and a reactive algorithm are investigated based on SWP.According to the simulation results,we find that there is a fundamental trade-off between the average content provisioning cost and edge cache hit ratio,and the proactive algorithm outperforms the reactive algorithm.