Researches On Optimal Deployment And Performance Evaluation Of Network Service Function Chain

With the rapid development and change of the Internet,the traditional network architecture has become increasingly unable to meet the needs of production and users.In the traditional network architecture,the relationship between the network Service Function(SF)and the dedicated hardware device is tightly coupled,so that network resources are difficult to share and new network services are difficult to integrate.As the scale of the network continues to expand,the traditional architecture has to invest more cost to deploy and operate new services.The concept of the network Service Function Chain(SFC)was born in the context of the rising of Network Function Virtualization(NFV)and the advancement of Software Defined Networks(SDN).A network service function chain is an end-to-end service traffic chain consisting of two or more Virtual Network Functions(VNFs).It aims to reduce network expansion costs and operating costs.It also has the ability to improve network resource utilization and speed up network and service deployment.This thesis will build a number of solutions to solve the problems and challenges of the current network service function chain.Researching and discussion on the development and problems of network service function chain,put forward the network service link inflow optimization model and algorithm under mobile network.Under the Industrial Internet put forward an SFC performance optimization model and algorithm based on SDN-NFV network and a network service function chain request prediction and optimization model based on wavelet neural network.This thesis first introduces the new network architecture SDN-NFV network architecture and lists the bottlenecks and challenges.Then the advantages and problems of the network service function chain are discussed from the IETF rules and transmission protocols.Among them,traffic,delay and energy consumption are the most concerned issues for users and network operators.Meanwhile these issues are also the key part of network performance.Therefore,this thesis establishes a set of traffic access optimization model for optimizing network access based on SDN-NFV network combined with integer linear programming for server load and traffic access problems.Based on this model,a maximum service request access method MSF is proposed.The MSF method can obtain the maximum access of the single-entry single-egress network request.At the same time,compared with the Advanced Heuristic Greedy(AHG)algorithm,the MSF method can achieve up to 2 times the service chain traffic access.At the same time,this thesis establishes a Virtual Machines(VMs)deployment model OVPMN that optimizes the delay and guarantees energy consumption for the performance problem of industrial Internet service chain.The OVPMN model uses probability theory and convex optimization theory to flexibly change the relative weights of energy consumption and delay according to user needs.The goal is to obtain the sum of optimized energy consumption and total delay.The OVPMN model algorithm is based on the posterior probability and K-Shortest Path(KSP)algorithm.Compared with the traditional greedy algorithm,the deployment cost is only 0.5 to 0.8 times higher than the greedy algorithm.Finally,this thesis uses wavelet neural network to continue to optimize the above model,and establishes a network service chain request traffic prediction and optimization model WNNPOM.WNNPOM can predict smooth network traffic changes and automatically adjust the weight values of the OVPMN model to flexibly orchestrate network service functions based on requests.WNNPOM could further reduces the deployment cost of OVPMN.