Research On Virtual Network Mapping Strategy In Wireless Networks

With the sprout of various late-model applications and the ever growing trend of mobile devices.applications set different demands on the aspects of netv ecurity,service quality and scalability of the underlying network.However it is difficult for the current chimney vertical network to adapt itself to the new internet businesses and the change of new flow models,and thus network reconfiguration must be carried out.Diversified network nodes in 2G,3G,4G and 5G not only lead to the differences in users' experience but also bring heavy load to network operation and maintenance(high-efficiency support is hard to realize).To some extent current radio access system is stiff,and Capital expense(CAPEX)and operating expense(OPEX)of network remain high.A new,flexible,high-efficient,open and green radio access network is in urgent need.In order to make the traditional Network infrastructure better integrate into future Virtualization Network Architecture,based on the Network Function Virtualization(NFV),and Software defined Network(SDN),Virtualized Interface Management Layer(VIML)of the ETSI architecture,this paper puts forward"A Mapping Framework for Resource Scalability Collaboration(RSCMF)".The VNFs orchestration scheme is studied for providing "customization" of businesses.The matching method of the logical slice is studied for speeding up instantiation of the new network slice.The RSCMF is studied for realizing the maximized benefit of mobile network operators.In order to speed up instantiation of the network slices,this paper proposes "An Identification Decision Tree Learning Model(IDTLM)" with packet classification methods.To speed up the matching process,aiming at the quantity problems of VNFs,IDTLM reduces dimension by proximal gradient descent method;at the same time,in view of the problem that the value of VNF' attributes range widely.Lagrangian is used to cluster.As online judge data and training data are inconsistent in distribution,it puts forward an online reinforcement learning method to improve IDTLM generalization performance,making it able to adapt to the transfer of different scenarios,improving the automation of business launches and operations.The online test of IDTLM show that the accuracy rate can reach 96%;meanwhile IDTLM system is of high recognition speed and low misjudgment rate,and thus system can adapt to the transfer of different scenarios.In order to reduce energy consumption,in view of existing situation where base stations are designed for peak loads,this paper proposes“Two-Step Mapping Algorithm(TSMA)".Aiming at the problem that single base station dormancy increases network delay,TSMA establishes resource mapping multi-objective optimization model for super densely-distributed environment.On the premise of guaranteeing users' QoS requirements,TSMA by studying overlapping update between base station mapping and resources mapping,finds out the approximately least number of base stations to provide services,and sleep the rest base stations,so as to achieve the goal of energy saving.The simulation results show that the energy-saving effects of TSMA are superior to the single base station round robin dormancy and compared with that TSMA is less possible to cause business delay.In order to meet different target demand for network management platforms in different network loads for different groups,based on the traditional multi-objective optimization model,RSCMF proposes "Wireless Heterogeneous Virtualization Resource Multi-Objective Mapping(WHVRMM)".As each subject focuses on different goals and wants its goals to be optimal,the results are barely satisfactory if traditional compromised multi-objective optimization algorithm is used.Based on the changes of network load,temporal difference of reinforcement learning method is applied to solve the dynamic weighting parameters in the model,and the results of differential evolution algorithm for dynamic multi-objective optimization algorithm of each target optimization are close to the single objective optimization algorithm.Simulation experiments show that WHVRMM maximizes the interests of all parties,and provides a scheme to achieve win-win results among multi-objectives.