Research On Joint Optimization Of Virtualized Network Function Chaining And Placement Problem In NFV

In order to improve network performance and ensure data security,strict planning and orderly execution of service requests are usually required in the form of service function chains.The emergence of Network Function Virtualization(NFV)technology has decoupled network services from hardware devices and realized the dynamic deployment of service function chains.NFV technology has greatly changed the way network service providers provide services to users.While NFV improves network management flexibility,it also reduces network Capital Expenditures(CAPEX)and Operational Expenditures(OPEX).A key issue in implementing NFV technology is how to effectively orchestrate and deploy Virtualized Network Function(VNF)sequences in order to allocate network resources reasonably and improve network performance.Since this problem is an NP-hard problem,solving this technical difficulty is full of challenges.At the same time,the chain composition of VNF is closely related to the deployment and affects each other,so the overall resource optimization cannot be realized only considering a certain stage.Therefore,this dissertation proposes a joint optimization strategy for VNF chain composition and deployment.The main contents are as follows:(1)Minimizing Operational Expenditure of Requests by Jointly Considering Virtualized Network Function Instances Composition and Placement: With the increasing number of service requests,network operating costs are becoming more expensive.However,due to the characteristics of VNFs that change traffic,the dependency between VNFs,and the complexity of composition and deployment methods,the composition and deployment of OPEX-oriented VNFs are full of challenges.To this end,this dissertation formulates the node mapping cost,link mapping cost,activation cost and energy consumption cost as OPEX,and establishes a new mixed integer nonlinear programming(MINLP)model.In order to analyze the different factors that affect the cost,this dissertation divides the VNF request set into completely disordered,partially ordered and completely ordered VNF request sets according to the dependency relationship,and designs three corresponding optimization algorithms.Experimental results show that the algorithm in the first two cases is better than other existing algorithms.For a fully ordered algorithm,when the ratio of node resources is above50%,an accurate solution to the approximate linear programming model in a small-scale network can be obtained.(2)Joint Optimization of Throughput and Operational Expenditure for VNF Chaining and Placement: For the ever-increasing network request flow,how to maximize network throughput under limited resource conditions and reduce network costs as much as possible is an important optimization direction for VNF chain composition and deployment.In response to this problem,this dissertation established a MINLP system model and designed two joint optimization algorithms for chain composition and deployment.The experiment evaluates the performance of the two algorithms under different resource constraints and the number of network requests.The experimental results verify the superiority of the algorithm in this dissertation compared with other comparison algorithms.Among them,the Genetic Algorithm(GA)algorithm can obtain a better throughput and cost joint optimization strategy,and the THRO algorithm can obtain a deployment plan in a shorter time.