Research On Multi Dimensional Resource Management In Mobile Edge Networks

In recent years,with the rapid increase in the number of intelligent user terminals and the emergence of many new applications,the traffic of mobile cellular networks has also increased exponentially.This trend results in high level of traffic load on the backhaul link and high transmission delay caused by congestion,which imposes huge challenge for the the traditional centralized network architecture.In response to this challenge,a new network architecture called mobile edge networks is proposed,which moves some of the network functions and remote content originally located at the core network to the edge of the network which is closer to end users.In addition to communication resources,computing and caching capabilities are added to the network,which form a new network architecture where multi-dimensional resources coexist.With regard to the new opportunities and challenges of mobile edge networks,this dissertation explores how to efficiently utilize multi-dimensional network resources such as communication,computing and caching to improve the service capabilities of future networks for diversified and differentiated service requests.It starts from the accurate modeling of the dynamic pattern of real traffic data,which lays the foundation for solving the mismatch between dynamic traffic and fixed resource management.By using this traffic model,a set of cooperative edge computing theories and methods for spatial non-uniform traffic are proposed.Under the principle of content centric allocation of cache resources,several cache resource allocation strategies are proposed as for the cache size limit of base stations and the selfishness of terminal users.The research conclusions provide new technical methods and theoretical support for the collaborative management of multi-dimensional resources of the next generation networks.The main contributions and innovations of this dissertation are as follows.1.Mobile traffic modeling method based on real network dataThe user traffic is dynamic and non-uniform both in the time domain and the space domain.The construction of traffic model in cellular networks based on real network traffic data can accurately describe the dynamic pattern of the traffic in actual networks,and solve the mismatch problem of dynamic traffic and fixed resource allocation.Thus,the resource utilization efficiency of the mobile edge network can be improved.The main innovative work of this part is to model the traffic patterns in multiple dimensions based on real traffic data of an existing mobile network.Firstly,as for the total traffic of multiple base stations in a region,a sinusoidal superposition model is proposed in the time domain to simulate the variation of the total traffic volume in an area.Through comparison of the fitting result of the proposed model and the real traffic data,it shows that the proposed model can accurately reflect the actual variety of the traffic and facilitate theoretical analysis.Secondly,as for the statistical distribution of the total traffic of multiple base stations,by modeling and analysis of the traffic data in several typical scenarios,results indicate that lognormal distribution can usually be used to model the statistical distribution of the network traffic,and the standard deviation parameter of the distribution model is related to the type of typical scenario,and the empirical values of the standard deviation parameters in the three typical scenarios are given respectively.Finally,a single base station traffic model for service scenarios is proposed.The model is capable of simultaneously characterizing the periodicity and randomness of the traffic variation of a single base station.The accuracy of the model is verified by comparing the data generated by the model with the traffic data of the real cellular network.2.Cooperative edge computing scheme under spatial non-uniform traffic distributionThe computing resource of a single edge computing platform is limited,and the traffic distribution in the edge network is heterogeneous and dynamic.This dissertation proposes a cooperative edge computing scheme that is suitable for these traffic characteristics and breaks the resource constraint of single platform.The scheme reduces the long term average energy consumption of mobile edge networks under latency constraints by task migration between edge computing servers deployed at each base station and sleep control mechanisms adapted to traffic tidal effects.The main innovative work of this part is as follows:Firstly,based on the traffic statistical distribution model,a traffic arrival model that can fully reflect the actual variation pattern of traffic is established,and the energy consumption and delay of the communication and computing resources of the edge network are modeled.The long term average energy consumption of the mobile edge network is optimized by jointly designing the sleep control mechanism and the computation task migration strategy of the edge computing servers.Since the direct calculation of the long term energy consumption requires future information,which is impossible in practice,an online algorithm based on the Lyapunov optimization method is proposed to make the migration decision in the current time slot without obtaining the future network information.By evaluating the system performance and analyzing the impact of system parameters,the results show that compared with the two baseline strategies:the one without computational cooperation and the one without sleep control,the proposed strategy can save more than 30%of energy consumption.At the same time,the proposed strategy maintains a relatively low latency compared to the baseline strategies.3.Caching schemes of base stations and devices in mobile edge networksAccording to the content centric network design principle,moving user requested content from a service provider to the edge of the network can significantly improve users' experience.In this dissertation,the content caching problems at base stations and at user devices in the edge of the network are studied respectively.The allocation strategy of cache resources is studied to maximize the network utility and reduce the cost of network services at the same time.The main innovative work of this part is as follows:Firstly,as for the selection of content placement in the distributed caching resources of base stations,the advantages and disadvantages of the two schemes of fully redundant caching and fully diversified caching are analyzed respectively.Then a caching strategy obtaining the tradeoff between the two schemes is proposed to minimize the total transmission cost of the network.The optimal redundancy ratio under a given system configuration is obtained by the particle swarm optimization algorithm,and the network can save more than 50%of the transmission cost under the optimal redundancy ratio.Secondly,by analyzing the impact of system parameters,it is shown that the optimal redundancy ratio is mainly affected by the unit transmission cost ratio of backhaul to RAN(Radio Access Network)and the steepness of file popularity.Finally,as for the selfishness of users,a joint design method of user incentive scheme based on contract theory and device caching is proposed to maximize the utility of each participant in the network and to achieve a win-win situation between the base station and user devices.After classifying the user devices according to the preference of sharing their cache resources,the base station pays corresponding incentives for different types of devices.Under the goal of maximizing the utility of the base station,an optimal contract is obtained through a heuristic caching algorithm.The research results show that compared with the two baseline schemes of fair caching and random caching,the proposed scheme can result in higher overall network utility through rewarding the user terminals which contribute more caching resources.