Study Of Dynamic Deployment And Readjustment Optimization Methods For Service Function Chains

With the emergence of Software Defined Network(SDN)and Network Function Virtualization(NFV),network operators can provide network functions to users in software form on general servers,which are also called Virtualized Network Function(VNF).Service Function Chain(SFC)refers to a set of ordered VNFs connected by virtual links established by SDN in the network to meet the diverse needs of its users.How to deploy service function chains in NFV/SDN-enabled networks is a problem that deserves in-depth research and is becoming a hot topic in academia,where some research has often struggled to balance deployment flexibility and success.To address this problem,this paper proposes a dynamic SFC deployment and adjustment method based on Deep Q Network(DQN)and M-Shortest-Path Algorithms.This method aims to maximize the request acceptance rate by modeling the dynamic deployment and readjustment of SFC problem.In the proposed method MQDR,we use two agents to dynamically coordinate the deployment and readjustment of SFC,so as to improve the service request acceptance rate.To reduce the action space and the training difficulty,we apply MSPA for dynamic deployment and reduce the action space for readjustment from two dimensions to one.Simulation experiments show that MQDR improves the request acceptance rate by about 25%compared to the original DQN algorithm and 9.3%compared to the Load Balancing Shortest Path algorithm.This paper addresses the dynamic embedding problem of SFC in dynamic traffic scenarios and proposes the algorithm MQDR+.The algorithm adds a traffic monitoring system to MQDR algorithm,which can timely select appropriate readjustment strategies according to network state and service request traffic changes,to optimize network resource utilization and service quality,and to avoid service interruption caused by dynamic traffic changes.Through simulation experiments,we compare and analyze the performance of MQDR+,MQDR and genetic algorithms.We show that in dynamic traffic scenarios,MQDR+algorithm improves the request completion rate by 11%compared with MQDR algorithm,while the decision time is only one tenth of genetic algorithm.We verify the effectiveness and flexibility of MQDR and MQDR+algorithms in dynamic deployment and readjustment of SFC through simulation experiments.This paper provides a new idea and method for dynamic management of SFC,which has certain value and significance.