Service-Oriented Resource Management In Software-Defined Wireless Networks

Recently,software-defined wireless network(SDWN)is regarded as a very potential newly-emerging architecture and attracts huge attention from both academia and industry.Through the decoupling of control and forwarding functions,SDWN can not only provide a more flexible,powerful and efficient resource management mechanism,but also realize new features and functionalities such as service-oriented resource management,artificial intelligence aided resource management and heterogeneous resource cooperation.Therefore,SDWN can meet the upcoming demands from more complicated application scenarios(e.g.,highdensity)and more versatile newly-emerging services(e.g.,VR/AR).However,SDWN is quite different from traditional wireless networks.Current research on resource management in SDWN is still in its infancy,and one key problem is how to design the resource management pattern that is appropriate for SDWN so as to fully utilize its features.To this end,this thesis studies service-oriented resource management in SDWN.First,this thesis studies the static wireless local area network(WLAN),and solves the client association management problem considering latency-sensitive services under the “client-infrastructure” architecture in software-defined WLAN(SDWLAN).We study the “Client Association in Software-Defined WLAN”problem.We first introduce how SDWLAN works and indicate the new features that affect client association,including: centralized client association,global network state awareness,flow-level association and seamless handoff.Next,based on the Markov analytical model of WLAN,we derive the average inter-packet of individual flows.With the above analytical results,we formulate the delayminimized client association problem.In order to solve the problem,the original problem is transformed as a supermodular set function minimization problem.A greedy algorithm and a bounded local search algorithm are developed correspondingly.Finally,with extensive simulation results,we analyze the influence of some main network factors on delay performance,and validate that our methods can obtain 20% to 75% lower latency than existing algorithms under different network environments.Then,this thesis studies the dynamic Internet of vehicles(IoV),and solves the dynamic access of white spectrum considering vehicular content delivery services under a similar architecture in software-defined IoV(SD-IoV).We study the problem of “Dynamic Access of White Spectrum in Software-Defined Internet of Vehicles”.We first introduce the network architecture and design issues.Then,we develop the resource model for spectrum-database-aided TV white space resource,as well as the microscopic mobility models for vehicular users.Based on the models,we formulate the resource allocation problem from two different perspectives and bring forth corresponding algorithms.The resource allocation optimization problem aims to maximize the total throughput.Here we propose a location-based resource allocation algorithm and a location-based maximal frame size design method,both of which can improve the total transmitted data of vehicular users.For the resource matching problem,we consider both throughput and users' QoS and prefer the latter.Through resource matching algorithm,stable and vehicle-optimal matching between available channels and vehicular users is realized.The advantages on throughput,QoS and fairness of our methods against baseline methods and comparison methods are validated via extensive simulation results.At last,this thesis takes into account the ad hoc nature of IoV and studies the resource allocation problem considering services with different QoS requirements in SD-IoV.We study the problem of “Service-Oriented Resource Management in Software-Defined Internet of Vehicles”.We first describe the system model and network architecture and give the workflow of resource management process.Afterwards,we formulate the service-oriented resource management problem as overlay vehicular network creation problem.We further design the service-oriented utility function for evaluating overlay vehicular network and use it as the optimization objective.Due to the complexity of the problem,we mainly adopt the graph-based genetic algorithm for modeling and solving the problem.The main procedures of genetic algorithm are re-designed,including: genotype representation,initial population generation,crossover,mutation and fitness evaluation.Besides,we also develop an easy-to-implement greedy algorithm.In the end,through extensive simulations,we analyze the effect of parameters on the performance of graph-based genetic algorithm,and validate that the proposed methods can achieve twice the total transmitted data and finished tasks than existing methods under high vehicle density environment.