Research On Novel NFV Resource Allocation Models And Algorithms

With the rapid development of novel network technologies and diversification of service requests,the scale of information networks is expanding,and many problems in traditional networks are becoming increasingly evident.Network Function Virtualization(NFV)as a new technology provides a way to solve these problems.NFV decouples network functions from proprietary hardware and serves users' requests by deploying Virtualized Network Functions(VNF)on common servers with high flexibility,high scalability,and low costs.In the NFV scenario,the Service Function Chain(SFC)connects a set of VNF in a certain order through virtual links to provide services for service requests.This dissertation mainly studies on the NFV resource allocation(NFV-RA)problem,with emphasis on studying the SFC mapping problems based on both the scenario with nonuniform users' demands and the VNF Resource Amount Change Matrix(RACM).Chapter one introduces the research background and value,and summarizes the current research status on NFV and SFC mapping.Chapter two introduces the NFV reference architecture and provides an overview of the NFV resource allocation problem,including the three phases of the problem,constraints and optimization objectives.Then it outlines the SFC mapping problem,compares the difference between SFC and VNF-forward graph.It introduces the principle and algorithm classification of SFC mapping,and concludes the research status of the SFC mapping problem in conjunction with the research direction of this dissertation.Due to the influence of geographical factors,population distribution and economic level,user service requests are not uniformly distributed in the network.Thus,the corresponding nonuniform allocation of network resources could meet user needs better.Chapter three focuses on the SFC mapping problem in the scenario where node resources and users' demands are nonuniform in NFV.In this thesis,it combines the node importance and ondemand service requests,formulates the SFC mapping as an Integer Linear Programming(ILP)model,and propose a Node-resource-and User-demand-Aware Mapping(NUAM)algorithm.Then it designs a VNF Reuse and Deployment Mechanism(VRDM)to implement the reuse of VNF instances and the reduction of bandwidth consumption.The simulations indicate that the proposed NUAM algorithm can reduce the bandwidth consumption and increase the acceptance ratio of requests significantly compared with current methods.The impact of VNF on the network is generally assumed to influence only the bandwidth,which is obviously far from the real application scenario.Thus,the tutor research group proposes the concept of VNF Resource Amount Change Matrix(RACM)for the first time.Chapter four focuses on investigating the SFC mapping problem based on the RACM of VNF.To address this problem,an ILP model is formulated with the optimization objectives of reducing both the node compute resource usage and the link bandwidth resource usage.This chapter proposes a Mapping algorithm based on Resource Amount Change Matrix(MRACM)to solve the model,which is able to reduce the link bandwidth usage based on the maximum reuse of VNF instances.In addition,a new VRDM based on greedy algorithm is designed on the basis of the work in Chapter three and combined with the RACM.The simulation results show that MRACM can achieve good results and verify its effectiveness.Chapter five summarizes the work in this dissertation and gives further work that can be developed in terms of network scenarios,SFC models,network resource types,resource amount change matrix of VNF and optimization algorithms.