Research On Distributed Caching Technology In Wireless Converged Networks

With the growing maturity of the Internet,the popularity of mobile intelligent terminals,the rapid development of broadband multimedia services,and the emergence of new communication scenarios such as immersive media,vehicular communications and interconnection of all things under the background of 5G,the global mobile data traffic shows exponential growth,which brings great challenges to wireless communication networks.Aiming at the problem of traffic overload in wireless communication network,the unloading of wireless fusion network based on cache is a good way to solve the problem.However,the centralized cache mechanism used in wireless convergence networks has the problems of limited throughput improvement and low cache utilization.In this paper,distributed caching technology in wireless fusion networks is studied.Firstly,the model of wireless fusion network system based on distributed caching is studied in this paper.For user-intensive scenarios,such as high-speed rail,railway stations,etc.,this paper introduces wireless Mesh network in wireless convergence network.To simplify the complexity of the problem,the Mesh routing node has a grid topology.Due to the limitation of node capacity and user request tolerance time,the system model only considers scenarios within two hops,that is,user requests to neighbor nodes.Then,in the case of uniformly distributed users,under the constraints of buffer and bandwidth based on nodes,the optimal service-to-node allocation strategy maximizes the throughput of the system and is modeled as an optimization problem.This paper proves that the problem is a non-polynomial time solvability(NPH)problem,and proposes three heuristic approximation algorithms: alternation based node-selecting and odd-even based dynamic programming algorithm(ABNS-ODP),alternation based node-selecting and dynamic programming algorithm(ABNS-DP),degree-based greedy-set-cover node-selecting and dynamic programming algorithm(DBGNS-DP).The simulation results show that the throughput of the distributed caching strategies implemented by the three algorithms is better than that of the centralized caching in the traditional fusion network,and the cache utilization is increased by %50 to %70.In terms of throughput index and cache utilization,DBGNS-DP has the best performance,followed by ABNS-DP and ABNS-ODP.In terms of complexity,the performance of the algorithm is opposite,but far less than the complexity of the optimal traversal solution,which shows the tradeoff between performance and complexity.In terms of request response time,the algorithms are all fixed values.In terms of load balancing,the performance improvement of DBGNS-DP algorithm is at the expense of load balancing,delay time and complexity.It needs to make a compromise between ET and load balancing,complexity and delay.Then,in view of the non-uniform user distribution scenario,the spatial homogeneous Poisson point process and Poisson cluster process are introduced to model the user request distribution.Based on the uniform distribution scenario and considering the load capacity problem caused by the non-uniform distribution of user requests and the bandwidth constraints of routers,requests-based greedy-set-cover node-selecting and dynamic programming algorithm(RBGNS-DP)based on the number of requests is proposed.The simulation results show that the performance of RBGNS-DP is better than that of traditional fusion network dynamic programming algorithm.For the general user non-uniform distribution model,the proposed algorithm can be used to solve the problem,which also shows the universality of the algorithm.Finally,the above research work is summarized and the future research is prospected.