Joint Optimization Of Network Function Placement And Traffic Scheduling In Service Function Chain Deployment

The proposition of network function virtualization(NFV)aims to solve the ossification and difficulty in current network's management and service provision caused by the deployment of numerous dedicated hardware devices.By decoupling the Network Functions(NFs)from dedicated hardware and deploying them in the form of Virtual Network Functions(VNFs)on common network devices,NFV realizes the separation of software and hardware and the flexible deployment of network functions and service function chains.However,the NFV has also brought a series of new challenges,including the service function chain deployment in NFV resource allocation.The service function chain deployment mainly involves three phases of optimization:the VNF selection phase,the VNF mapping phase,and the VNF inter-traffic scheduling phase.An optimal service function chain deployment should jointly and comprehensively consider above three phases as the three phases mutually affect and interact.However,most of current related researches only consider one of two phases' optimization of service function chain deployment and treat other stages as known quantities or inputs.Only a handful of works have studied the three-phase coordinated service function chain deployment problem,but the coordination performance is not very good.Hence,it is necessary to conduct a deeper study on the three-phases coordinately optimized service function chain deployment problem.In this paper,the VNF selection phase and the VNF mapping phase are collectively called as network function placement,and the VNF inter-traffic scheduling phase is called traffic scheduling,and a three-phases coordinately optimized service function chain deployment problem is studied,that is,joint optimization of network function placement and traffic scheduling in service function chain deployment.In this paper,a typical joint optimization model for network function placement and traffic scheduling is established first.Secondly,a modeling method called Homogeneous Link Mapping(HLM)is proposed and a CPLEX model which is corresponding to the above model is established with HLM,in order to use CPLEX optimization tool to obtain the optimal service function chain deployment schemes which are further used to evaluate the following designed heuristic algorithm.Then,this paper analyzes the impact of the substrate network with topology structures on service function chain deployment performance under current model.Finally,in order to realize quick arrangement and placement of service function chain,this paper designs a heuristic service function chain deployment algorithm called Merge-Split Viterbi(MSV),and analyzes the performance of the MSV algorithm in detail.