SFC Placement And Scheduling Problem In A Hybrid Senario

As the number of Internet users continues expanding,the rich application scenarios of large traffic,such as short video,live broadcast,have led to a significant increase in access traffic.With the popularization of big data,industries such as cloud computing and artificial intelligence have accelerated their development.These emerging businesses and network requirements are increasing,which intensifies the pressure on network providers in deployment,operation and maintenance,and resource allocation.It is well known that traditional proprietary hardware network functions(PNFs)are the essential elements to serve traditional network services,which we call middle-boxes.Generally speaking,middle-boxes are the forwarding or processing devices that transmit,transform,filter,check,or control network traffic,to achieve network control and management.Before applying them into use,the middle-boxes must be integrated into the network infrastructure through a complex deployment by specialized technicians,which requires expensive labor cost and time cost.The middle-boxes are fixed somewhere in the network and cannot be easily moved or shared.The network is becoming more and more ossified.In order to solve such challenges brought by middle-boxes,unremitting efforts and researches finally contributed to the emergence of network function virtualization(NFV).The network function virtualization provides users with flexible user experience,meeting the requirements of diversified businesses,while ensuring security and network manageability.However,it is impossible to completely transition to network function virtualization in a short term.In other words,the coexistence of proprietary hardware network function(PNF)and virtual network function(VNF)is inevitable.In this paper,we focus on the deployment strategy of service function chain(SFC)in VNF/PNF hybrid scenario in the optical data center network(ODCN).According to the personalized user demands,the optimal deployment strategy of SFC is searched as follows:(1)In order to meet the strict delay requirements of users,we propose a heuristic algorithm aiming at minimizing the total cost of VNF and PNF.In the research of low-cost SFC deployment strategy in VNF/PNF hybrid scenario,we design the principle of network function mapping,and define the basic time of the user demand,the processing capacity gap between VNF and PNF,etc.In the network function mapping principle,the number of PNFs and VNFs that serve a demand user and which network functions are preferentially mapped to PNF are specified in the network function mapping principle.(2)In order to better balance performance optimization and cost,considering that different users focus their attention on different factors,we propose a service function chain deployment strategy with a trade-off of total cost and delay in VNF/PNF hybrid scenario.In the perspective of users,the objective function focuses on the weighted values of cost and delay.We construct a mixed integer linear programming(MILP)model,and then propose a heuristic algorithm to overcome the scalability limitation of the mixed integer linear programming solutions.SFCs are mapped to VNF and PNF based on resource occupation,delay and cost consumption etc.,so as to truly integrating VNF with PNF.In order to simplify this research strategy,we assume all network functions can be virtualized in our scheme and experiment,corresponding PNF exists in the network,and the virtualization technology used is Virtual Machine(VM)only.Traffic data of the same demand must be executed in SFC sequence.Data for different user demands is executed serially in the same VNF instance and in parallel in the same PNF.Our algorithm ensure the total data size executed simultaneously in a PNF is smaller than the throughput of the PNF.