Research On Performance Monitoring And Optimization Of Virtual Network Function

Network Function Virtualization(NFV)implements network functions in software form and deploys them on general-purpose servers,alleviating the problems of inefficient deployment and high operating costs of traditional dedicated equipment.However,performance is the key factor that restricts the wide use of NFV.Most of the existing NFV performance research methods do not consider the interference of the juxtaposed Virtual Network Function(VNF)on the running Service Function Chain(SFC).The application of these methods to SFC performance monitoring,VNF deployment and migration optimization often leads to performance monitoring errors,low resource utilization and low migration success rate.Therefore,this thesis focuses on the monitoring and optimization of VNF performance to improve the accuracy of SFC performance monitoring,network resource utilization and SFC service availability.The main work is as follows.Firstly,to address the problem that traditional performance monitoring methods cannot accurately perceive SFC performance,this thesis proposes a novel VNFs performance monitoring model by considering resource competition between juxtaposed VNFs,physical server load and other factors.When resource competition is formed,the resources are first reallocated through the principle of fair competition;then the server load influence model is used to calculate the load influence factor and combine the allocable physical resources to obtain the expected SFC performance;finally,the proposed model is applied to the SFC mapping.The experiments show that the model can effectively monitor SFC performance.Secondly,to address the problem that the current SFC scheduling method does not take into account the performance interference of VNF juxtaposition,which leads to the low acceptance rate of subsequent SFCs,an Optimized Genetic Algorithm(OGA)is proposed in this thesis for SFC scheduling.The algorithm is based on the VNFs performance monitoring model and takes into account the performance impact of VNF juxtaposition while minimizing deployment resource consumption in order to improve the resource utilization of the physical network.First,OGA uses the VNFs performance monitoring model to calculate the impact of new user requests on deployed services;then the magnitude of the impact is used as the weight of roulette in the genetic algorithm to guide population evolution;and finally,the optimal solution is transformed into an SFC scheduling scheme.The experiments show that OGA is significantly effective in improving request acceptance rate and resource utilization.Finally,this thesis proposes a highly available VNF migration optimization algorithm to address the problem that the current VNF migration method does not consider the overall performance of SFC during and after migration,which results in a low migration success rate.First,the VNFs performance monitoring model is used to calculate the post-migration performance impact factor;then the migration overhead is calculated using the business flow of the migrated VNF;then the post-migration performance impact factor is used to calculate the performance achieved after the migration;and finally,the migration solution with the least user service impact is selected for the migration by combining the migration overhead and the post-migration performance calculation.Experiments show that the proposed algorithm in this thesis can effectively solve the problem of severe degradation of subscriber service quality in the migration optimization phase of VNF.