Research On The Mapping Of Service Function Chains In Network Function Virtualization Scenario

With the rapid development of the Internet technology and the ever-emerging of new services,it is more and more difficult for the large and rigid telecommunications networks existed to satisfy the ever-changing user requirements.Network operators are eager to reduce network complexity and speed up the deployment and application of new services.Network operators with network function virtualization(NFV)technology can realize virtualized network functions(VNF)by using industry-standard servers,storage,and switching equipment.VNFs can be used to replace expensive and dedicated network elements that are difficult to be maintained and upgraded.New services can be developed and deployed quickly and flexibly by using VNF.Therefore,NFV can accelerate service innovation,and reduce operator's capital expenses and operating expenses.Network services usually need to be processed through various network equipment,such as firewalls and load balancers.The Service Function Chain(SFC)describes service requests by defining such a network functions set that its elements are ordinal or partially ordered.The research on the deployment of SFC in NFV networks has great theoretical significance and practical prospect.On the basis of discussing the status of SFC mapping research,this thesis focuses on both the SFC mapping problem in which multiple VNFs are dependent and the SFC mapping problem in hybrid NFV scenarios.The first chapter of this thesis introduces its background and significance and then summarizes the research status at home and abroad.The second chapter mainly addresses the research basis of the SFC mapping problem in the NFV scenario.The reference architectural framework of NFV and functions of its modules are summarized.Then,the mapping problem of SFC is introduced with the analysis on the relationship between SFC and VNF Forwarding Graph(VNF-FG).We also compare the similarities and differences of the SFC mapping and virtual network mapping,and then the existing SFC mapping algorithms are summarized based on different classification criteria.Finally,the multi-objective optimization problem and the fundament of the multi-objective evolutionary algorithm are described briefly.For the SFC mapping problem in which multiple VNFs are dependent,the third chapter focuses on dealing with two stages of building SFC and mapping SFC jointly.Firstly,an integer linear programming(ILP)model of SFC mapping in Elastic Optical Networks(EONs)is established.Then,a coordinated SFC building and mapping algorithm is proposed and implemented for reducing physical resource consumption.Compared with the existing algorithms,simulation experiments show that the proposed algorithm could decrease both the maximum frequency slot index and the total number of VNFs deployed in EONs,and reduce the resources consumption of physical network effectively.Due to economic factors and the gradual employment of new technologies and equipment,equipment of two types with the dedicated network function and VNFs will coexist in telecommunication networks for a long time.The fourth chapter concentrates on studying the SFC mapping problem in hybrid NFV scenarios.Aiming at reducing both the server computing resource consumption and the link bandwidth resource consumption,a bi-objective SFC mapping model is established in hybrid NFV scenario.Then,on the basis of the multi-objective evolutionary algorithm with decomposition(MOEA/D),a bi-objective evolutionary SFC mapping algorithm based on decomposition(BOESMA/D)is proposed and implemented to solve this model.To improve BOESMA/D algorithm's performance,a mapping strategy based on SFC decomposition is designed to generate a portion of the initial population.Simulation experiments show that the proposed algorithm could effectively decrease the consumption of link bandwidth resources when it is compared with an existing algorithm.In addition,compared with the randomly initialized population,the proposed population initialization method could accelerate the convergence of the proposed algorithm and increase its diversity of solution sets.The fifth chapter summarizes the works of this thesis and presents the following five future study issues,including: service request model,SFC model,coordinated algorithm in three-stage,cost model,SFC mapping model in hybrid NFV scenes.