Research Of Service Function Chain Embedding Under A Multi-domain NFV Environment

With the Network Function Virtualization(NFV)technology,network device is decoupled as software and hardware parts.The software of network function then can be hosted by cheap general computation devices rather than costly dedicated middleboxes.To obtain a specific network service,the traffic flow is steered to go through a specific order of network functions,and this network function sequence is known as a service function chain(SFC).In the environment of NFV,the SFC provides a flexible way for service providers to easily adjust their functions and expand their business.Under the environment of NFV,the bottom layer provides a unified resource pool,and cloud service providers provide a variety of service functions.By managing SFC through SDN,the deployment and management of SFC can be easily realized.The advantages of flexible deployment and automatic expansion of service function chain make it the focus of attention of academia and industry in recent years.Nowadays,multi-domain network in multi-service provider environment poses new challenges to the deployment of service function chain.To deploy service function in multi-domain environment,the allocation of computing node resources and link bandwidth resources shall be take into consideration,as well as the information isolation brought by multi-domain environment.Therefore,how to optimize the deployment of service function chain between domains and how to deploy resources in service function chain has reached a win-win situation for both users and operators.In the thesis,the author firstly studied the impact of multi-domain environment on traditional SFC embedding.In chapter 3,the author proposed a chain-cutting hierarchical embedding algorithm and a heuristic algorithm based on multi-domain spreading.The author studied and analyzed the impact for SFC embedding under different level of imformation isolation.In Chapter 4,the author proposed multi-domain embedding approach based on column generation algorithm.The original embedding problem can be decomposed into two parts: inter-domain mapping and intra-domain mapping CG algorithm and can be solved iteratively by RMP and PP.An optimal solution can be obtained eventually.Considering the time-consuming in CG,the author proposed a Rounding Off algorithm to simply branch-and-bound a heuristic algorithm based on auxiliary graphs to accelerate PP.The simulation results show that the performance of the algorithm is greatly improved while the embedding results remain approximately optimal.Finally,in Chapter 5,the author adds the delay constraints to the multi-domain SFC embedding problem,and takes the dynamic resource allocation into considertaion.According to the new constraints,RMP and PP models in column generation algorithm are readjusted to minimize SFC embedding cost with guaranteed delay.