Research On Key Technologies Of Software Defined WLAN System

Currently,Wireless Local Area Networks(WLANs)are employed as an indispensable part of network infrastructure and various application scenarios are constantly being expanded and developed.However,the scalability and flexibility of the system are poor due to the highly tight-coupled architecture between hardware and software.Software Defined Network(SDN)and Network Function Virtualization(NFV)have been introduced into Software Defined Wireless Network(SDWN)which can change the rigid of traditional system and enhance the flexibility,programmability and scalability.Although the mainstream SDWN systems architecture can greatly reduce the coupling between physical hardware and logical applications,the challenges of high-speed forwarding in Data Plane and the precise decision-making and efficient management in Control Plane are critical when the network scale becomes larger.To deal with these challenges,three aspects are studied.Firstly,network awareness is the premise for the Controller to realize precise decision-making.A low-cost network measurement model based on Matrix Completion(MC)algorithm is designed under limited wireless resources.Secondly,although virtualization technology improves network programmability,it introduces additional packet processing delays that causes serious performance issues.This dissertation designs a fast packet forwarding method to reduce the processing delay.Thirdly,in the high-density deployment scenarios,energy consumption and interference have become serious.A joint optimization scheme of energy efficiency and interference is designed which can minimize system energy consumption and reduce interference simultaneously.The main works are shown as follows:(1)Research on a low-cost network measurement model.The Controller needs to accurately measure the network to make reliable decisions in real-time.However,the Access Point(AP)has poor hardware resource which cannot afford the measurement overhead for all nodes.In order to minimize the cost of network measure,an online real-time measurement method based on MC algorithm is proposed.The complete matrix can be reconstructed from partial measurement samples with low error rate.However,the traditional MC schemes are designed for offline application scenarios and usually assume that the data matrix has a known and fixed low-rank.To handle this problem,an online adaptive MC algorithm based on Random Walk(RW)is proposed.Firstly,based on the observations of real network measurement trace,the features of low-rank,temporal stability and relative rank stability are found.Secondly,since the rank of matrix is not fixed and changes dynamically with the wireless environment,this dissertation proposes a sliding window-based measurement model.Meanwhile,in order to satisfy the precision requirement of matrix reconstruction,the RW model is introduced to provide a guide to the node samples at new time slot.Finally,to realize the dynamic adaptive selection of measurement points,the sampling node sequence at new slot is dynamically adjusted according to the reconstruction error between adjacent time slots.The experimental results demonstrate that the proposed scheme can reconstruct the complete matrix with required precision by only a small sampling rate about 30% and the reconstruction error is only 0.6%?0.7%.(2)Research on the high-performance AP virtualization design.The application of virtualization technology realizes the decoupling between hardware and software by constructing the logical relationship between terminal and virtual BSSID which effectively improves the programmability and flexibility.However,in the existing SDWN systems,all wireless frames are extracted into User space by listening interface mode which leads the degradation of VAP forwarding performance due to the frequent context switches,large memory copies and system calls.The forwarding efficiency of Data Plane drops sharply.Therefore,for single AP virtualization,a new AP virtualization architecture is proposed which combine User space and Kernel space to decouple wireless frames.Secondly,to ensure the high mobility of the terminal,a method for cooperatively maintaining terminal access information between the AP and the Controller is designed.When the terminal performs handover,the Controller sends the message to the neighbor APs.Thirdly,throughput performance and robustness are further improved by introducing a multi-connection mechanism.To maximize each application's throughput,the flow assignment is formulated as a Mixed Integer Non-Linear Programming(MINLP)problem.A heuristic algorithm with low complexity is proposed which determines a feasible optimal solution by narrowing search set with cutting-off solution space,namely,NS-Co S.Finally,delay and throughput performance are verified in single connection and multiple access scenarios respectively.(3)Research on a joint optimization of energy efficiency and interference.Aiming to solve the problem of energy waste and interference caused by densely deployed APs,traditional methods usually consider them unilaterally and ignore their mutual influence which lead to the failure of optimal solution.By making full use of the resource dynamic scheduling strategy of AP virtualization and the Controller's global view,a joint optimization scheme is proposed.During the optimization process,the dormant AP set is selected and the transmitted power of working AP is dynamically adjusted.Firstly,through the quantitative measurement and analysis of the energy consumption of general AP equipment,the mathematical model of throughput,transmit power with energy consumption is constructed based on the experimental results.Secondly,due to the limitation of wireless resources,a Bayesian based mechanism is proposed with joint optimization on energy efficiency and interference.The social selection function is used to solve the optimal dormant AP set and power adjustment rules.Moreover,when there are enormous network nodes,the iterative-based fast algorithm is proposed by using clustering and iterative ideas.Finally,simulation results shown that the optimization algorithm proposed can effectively reduce energy consumption and interference.The low-overhead network measurement is achieved based on MC and RW and the high-performance virtualization architecture is designed.By decoupling wireless frames and zero-copy,these schemes can effectively support the Software Defined WLAN system with flexible programming,resource scheduling and management.In order to verify the key technologies proposed in this dissertation,a prototype system has been designed and developed which only need software upgrade on AP and leave users' devices unchanged.