Research On Content Deployment And Resource Management For Wireless Distributed Caching

Wireless caching technology can achieve efficient traffic offloading by caching data at the edge of the network.Thus,it is one of the key technologies to solve the rapid growth of network data under the current popularity of the fifth generation(5G)networks.By reasonably deploying caching nodes at the edge of the network to cache popular content,it can not only reduce network burdens and decrease system energy consumption,but also reduce content transmission delays and improve user experience as well.However,due to the constraints on the caching-computing-communication capabilities of edge caching nodes,the benefits of content caching will be seriously affected.At the same time,the massive data interaction between edge caching nodes will consume a lot of spectrum resources,which will not only make wireless resources even more scarce,but will also bring huge additional energy consumption.Therefore,how to proceed from user needs and consider multi-domain resources,such as content distribution characteristics,user social relationships,caching node capabilities,physical channel conditions,etc.,to design and implement an edge caching system with high energy efficiency,high spectrum efficiency,and low latency to provide users with high-quality content sharing service is essential to this paper.The key technology and implementation of wireless distributed caching system are discussed from the point view of combining with theory and practice use in this paper.Firstly,we study the problem of physical-social cross-domain caching based content popularity prediction.Next,with the goal of minimizing energy consumption,the joint optimization of content caching and delivery is studied.After that,considering the physical channel condition,the wireless resources are optimized with the goal of maximizing the successful content delivery rate.Finally,the wireless distributed caching system architecture is designed and implemented by combining with the studies mentioned above.The main work and research results of this paper are as follows:1.Physical-social cross-domain caching-based content popularity prediction methodIn order to improve the effectiveness of content caching,it is necessary to predict the popularity of the content in advance.Considering the social relationship between users,a common interest model(CIM)based on Dirichlet distribution is proposed to predict content popularity.After that,the successful probability of content delivery based on this model is deduced by combining with the quality of physical communication links and the caching strategy of content in nodes.Based on the successful probability,the target problem is formulated as a posterior probability maximizing problem about content popularity and caching strategy.In order to solve the complex computing task of the objective function,this paper proposes a Gibbs sampling based machine learning algorithm to estimate the parameters of the CIM model,thereby transforming the complex computing task into a simple statistical calculation problem,which greatly reduces the computational complexity.The numerical results show that,comparing with the traditional Zipf fitting method and statistical probability method,the proposed method can increase the average success probability of content delivery by 7.6%under the premise of ensuring QoS.2.Joint strategy optimization of content caching and delivery for energy minimizationEfficient content caching and distribution strategy is the guarantee for wireless distributed caching system to provide users with high-quality content services.In order to effectively mine and utilize the caching and computing resources of user equipment,this research proposes a scheme that absorbs users'equipment with the function of Device-to-Device Communication(D2D)as the caching node of the wireless distributed caching system.In this scheme,we comprehensively consider the problem of caching node selection,content caching,and content delivery.First,by comprehensively considered the social relationship between D2D users and the quality of the physical communication links,a user effective network centrality estimation algorithm based on the PageRank architecture is proposed to solve the problem of caching node selection.Then,by comprehensively consider the user's caching space,physical communication link status,and content popularity,a belief propagation framework based distributed algorithm is proposed to optimize content caching strategy.After that,combining with the status of content caching,the quality of physical communication link between users,and the content request situation,another belief propagation framework based distributed algorithm is proposed to optimize the content delivery strategy.Finally,in order to further improve the energy efficiency of the system,a comprehensive optimization scheme based on the heuristic algorithm is designed.Also,we prove the effectiveness of the algorithm through computational complexity analysis and numerical verification experiments.Comparing with the traditional matching method,the proposed comprehensive optimization method increases the computing speed by 25%at the expense of a very small amount of performance.3.Joint optimization for power control and link scheduling under statistical CSIResource optimization can significantly improve the performance of the wireless distributed caching system,thus reduce system costs and improve service quality.In order to solve the problem of spectrum resource allocation in the wireless distributed caching system,we consider a scenario where D2D users reuse uplink frequency resources of cellular users in an orthogonal frequency division multiple access(OFDMA)supported cellular network.With the goal of maximizing the system transmission rate while considering the energy consumption optimization and link scheduling problems,the target problem is modeled as a mixed integer nonlinear programming problem by using the statistical channel state information(CSI).To solve the problem of spectrum interference between users,in the objective function optimization process,we first transform the problem into a fractional programming problem by using the Lagrangian dual method,and then the fractional programming problem is transformed into a quadratic problem by the means of convex optimization tools.Further based on the above conversion,we propose a comprehensive power control and link scheduling scheme,which can be solved by three optimization method,namely continuous optimization method,discrete segmented approximation method,and traditional discrete matching method,to maximize the system transmission rate while solving the interference cancellation problem.The numerical results show that,comparing with the continuous optimization method,the discrete segmented approximation method increases the computing speed by 41.2%with a 12.1%total average rate loss of the system.While,comparing with the traditional discrete matching method,the discrete segmented approximation method increases the overall average rate by 9.5 times with a 6.3%computing speed loss.In addition,by effectively scheduling links in different time slots,the solution proposed in this paper can also effectively guarantee the communication needs of edge users or users with poor communication quality.4.The architecture design and implementation of the wireless distributed caching systemIn order to test the effectiveness of the above three optimization schemes and apply them to actual environments,we propose a wireless distributed caching system architecture by comprehensively considering the D2D technology,the network slicing technology,the network function virtualization technology(NFV),the multi-access edge computing(MEC)technology,and the software defined network(SDN)technology.The implementation of the wireless distribution caching system in a practical environment is given as well.The proposed architecture is a two-tier service system solution.At the upper layer,we use the SDN controller to achieve comprehensive optimization and management of the network.At the lower layer,we use the SDN agent to achieve precise optimization and management of the local network based on local needs.In addition,by adding a distributed computing architecture,the SDN agent can also complete collaboration tasks spontaneously or with the assistance of the SDN controller.In order to further improve the effectiveness of the service of content sharing of the wireless distribution caching system,we add additional equipment management module and routing management module to the system to treat all nodes of the network as network equipment.By this method,we successfully realize the shielding of Internet Protocol(IP)addressing in traditional networks,thus building a content service-centric service network.In the process of system implementation,we do the development work based on embedded technology and application program interface(API)technology,and provide standard management and data interfaces for different network services.So that the system has good compatibility and openness,and at the same time,it can be embedded,extensible,programmable,and so on.Finally,in order to prove the feasibility and effectiveness of the system,we conducted experimental tests on the system.Comparing with the existing system,the numerical results show that this system is with good performance in terms of node access time,transmission rate,delay,jitter,throughput,and average response time.It also verifies the effectiveness of the content popularity prediction method,the joint optimization method of content caching and delivery,and the joint optimization method of power control and link scheduling,which has a good application prospect in wireless distributed caching system.