Research On The Reliability Design Of Network Function Virtualization

In recent years,with the development of Software Defined Networking(SDN),cloud computing and virtualization technology,Network Function Virtualization(NFV)has become a research hotspot.NFV is a potential technology that decouples network functions from expensive dedicated hardwares and implements them with software on general servers.NFV technology can significantly reduce the telecom operators' capital expenditures(CAPEX)and operating expenditures(OPEX),shorten the deployment and delivery time of new services,and significantly increase the flexibility of traditional telecom networks.However,NFV technology also faces challenges.Compared with carrier-class equipments,the reliability of general servers is lower.Therefore,this thesis deeply studies the reliability of NFV for different application scenarios and optimization objectives.First,in order to improve the reliability of service requests,the third chapter proposes a dedicated protection mechanism,which is consist of the dedicated-all-nodes protection mechanism(DANPM)and the dedicated-single-node protection mechanism(DSNPM).However,while improving the reliability of service requests,the cost of mapping will also increases.In order to solve the problem of minimizing the mapping cost with the reliability guaranteed of the service requests required,this chapter establishes two integer programming models,and designs DANPM algorithm and DSNPM algorithm.The simulation results show that DSNPM algorithm is better than DANPM algorithm in term of computing resource,bandwidth resource and mapping cost.Second,compared with the third chapter,the fourth chapter proposes a shared protection mechanism,which is consist of the shared-all-nodes protection mechanism(SANPM)and the shared-single-node protection mechanism(SANPM).However,while improving the reliability of service requests,the cost of mapping will also increases.In order to solve the problem of minimizing the mapping cost with the reliability guaranteed of the service requests required,this chapter establishes two integer programming models,and designs SANPM algorithm and SSNPM algorithm.The simulation results show that SSNPM algorithm is better than the other algorithms in term of computing resource,bandwidth resource and mapping cost.Finally,because the cascading failure phenomena exists in the NFV environment,the fifth chapter studies the cascading failure assessment problem,transforms them into a set-cover problem and solves it with an evaluation algorithm based on a greedy set-cover algorithm.At the same time,the deployment of service function chains to reduce the impact of cascading failures is studied.The simulated annealing deployment algorithm is proposed.Simulation results show that the simulated annealing deployment algorithm can effectively reduce the impact of cascading failures,and its performance is obviously better than the shortest path deployment algorithm.