Research On Caching Technology For Mobile Edge Networks

In order to cope with the explosive growth of mobile data traffic and the increasingly stringent performance requirements of new applications in the fu-ture,the ultra-dense deployment of base stations is foreseen as a key enabler for the fifth generation mobile communication(5G)networks.However,with the ultra-dense deployment of base stations,the increasing mobile data traffic will exerts a great burden on backhaul links which will become the bottleneck of the 5G networks.By deploying storage resource at network edge(e.g.,small base stations and mobile devices)to serve users with short distance,mobile edge caching technology can alleviate the burden on backhaul links and improve the network performance efficiently.However,there are several challenges in the study of mobile edge caching technology:(1)The decisions of edge caching placement and content delivery are strongly coupled;(2)It is hard to obtain the popularity distribution in practice;(3)The environment in mobile edge net-works is highly dynamic;(4)The requested content types are various in the mobile edge networks.Therefore,it is critical to study the joint optimization of edge caching placement and content delivery for different content types in highly dynamic networks with unknown popularity.This thesis mainly inves-tigates the caching technology in different scenarios of mobile edge networks to solve the above problems.The contribution and novelties of this thesis are summarized as follows:Firstly,the joint optimization algorithm of multi-timescale edge service caching placement and uplink resource allocation is proposed in the case of single cell to cope with the delay of service installation and configuration,the finite memory bandwidth of the edge server and the strong coupling of ser-vice caching placement and uplink resource allocation.The proposed algorithm solves the problem of unknown popularity in an adaptive way.In specific,the joint optimization of edge service caching placement and uplink resource allo-cation is modeled as a stochastic optimization problem to maximize the time-average network throughput.The Lyapunov stochastic technique is employed to decouple the formulated problem over time.For the decoupled determin-istic optimization problem,the service caching placement problem in the large timescale is transformed into a multi-dimensional knapsack problem and solved by using the dynamic programming method;the resource allocation problem in the small timescale is solved in a closed form.Furthermore,the asymptotic op-timality is proved.Simulation results demonstrates that the proposed algorithm can effectively improve the network throughput while ensuring the network sta-bility.Secondly,the joint optimization algorithm of edge file caching placement and user association is proposed in a multi-cell and multicast-enabled case to solve the problems of user association in overlapping areas and the strong cou-pling of edge caching placement and user association.The proposed algorithm solves the problem of unknown popularity in an reactive way.In specific,the joint optimization of edge caching placement and user association is formu-lated as a nonlinear combinatorial optimization problem to minimize the net-work power consumption.Then,the formulated problem is decoupled into two subproblems:the user association problem and the cooperative caching place-ment problem.For the user association problem,by introducing the definitions of user group and user cluster,the user association problem is transformed to the set partition problem and solved based on the genetic algorithm.For the co-operative caching placement problem,a algorithm based on the relaxed lower bound is developed.Simulation results show that both the power consumption and load balancing is improved by the proposed algorithm.Thirdly,the joint optimization algorithm of multi-timescale edge file caching placement and multicast scheduling is proposed in a multi-cell case to cope with finite communication resource and dynamic network environment.The proposed algorithm solves the problem of unknown popularity in an adaptive way.In specific,the time-average system cost is minimized by formulating the joint optimization problem of multi-timescale edge file caching placement and multicast scheduling as a stochastic optimization model.In specific,the single-timescale and multi-timescale adaptive optimization algorithms are proposed by exploiting the stochastic gradient descent method and the T-slot Lyapunov stochastic optimization method,respectively.The performance loss of the two algorithms is compared theoretically.Simulation results demonstrate that the proposed algorithm outperforms the random caching scheme and the caching scheme based on file popularity.It further validates the necessity of joint opti-mization of edge caching placement and content delivery.Finally,a social-driven edge file caching placement and resource schedul-ing scheme is developed in a multi-cell case to cope with the strong mobility of users in mobile edge networks and the small overlapping areas of small base stations.The proposed algorithm solves the problem of unknown popularity by predicting it.In order to address the popularity prediction problem in mobile edge networks,a double-layer network is constructed which consists of social layer and physical layer.The spreading process of files in the social layer is modeled as the epidemic model.Then,a low-complexity popularity predic-tion algorithm is proposed based on the discrete-time Markov chain.Based on the predicted file popularity,a heuristic edge caching placement and resource scheduling algorithm is proposed.Simulation results show that the developed algorithm can greatly reduce the system latency.