Distributed Middlebox Placement Based On Potential Game

Software Defined Network(SDN)and Network Fuction Virtrue(NFV)As two new types of networking technologies,scholars have done a lot of research on their basic architecture and federated architecture.With the development of these two technologies,the construction of the service chain takes a high degree of flexibility,and also takes the development of software-defined middlebox.In this paper,we study how to optimize the performance of the service chain,mainly complete the two aspects of the work.Based on the SDN-NFV architecture,the unified control architecture of joint control middlebox deployment and traffic forwarding is proposed.Second,the theory of game theory is introduced,and the potential game model is used to solve the problem of middlebox deployment.The main work of this paper is as follows:First of all,it introduces the knowledge of game theory and gives the reader an intuitive understanding.And discusses the feasibility of game theory applied to the deployment of middlebox,summed up the steps of game theory analysis.Second,for SDN and NFV,elaborate its basic architecture,features and functions,and study the advantages of integrating both.In order to unify control strategy-aware middlebox deployment and efficient traffic routing,we designed a unified control architecture.The architecture includes information flow and control flow,information flow collection and transmission strategy information,network function status information,topology information and resource information;control flow control network function deployment and traffic forwarding behavior.Finally,where the middlebox is placement in the network layout to optimize service chain performance.The performance metrics studied in this paper include the end-to-end delay and the bandwidth consumption.Using the graph matching method,the formal expression of the research index is deduced.Then,the problem of middlebox deployment is established by using the potential game model,and a distributed algorithm based on this model is given.Simulation of a large number of simulation scenarios shows that our proposed algorithm is very close to the optimal layout,and compared to the random layout,end-to-end delay optimized by 34%,bandwidth consumption optimized by 17%.Farther,the algorithm complexity improves greatly.