Research On Multi-controller Placement Problem In Software Defined Networks For Latency And Reliability Optimization

Software Defined Network(SDN)architecture separates forwarding plane of the from control plane in networks so as to simplify the design and management of the networks.The control plane in the SDN architecture consists of controllers,which are responsible for handling requests from the switches.Studies have shown that a single controller can meet the business needs of small networks.However,considering the scalability,reliability,etc.of the networks,a single controller is not sufficient to meet the demand.A logical centralized and physically distributed multiple controllers architecture is often adopted,which casuses a new problem about how to place the controllers and map the switches to the placed controllers.Previous reseach has shown that reasonable placement of controllers can improve network performance,such as reducing latency and improving reliability.In this thesis,we study the multi-controller placement problem to improve the delay and reliability in software-defined networks.The main research contents are as follows:(1)Research on reliability-aware multi-controller placement problems for improved latency.A primary path is to be used for each switch and controller pair when there is no link failure in the nework.When a link on the primary path fails,a backup path is established to ensure the switch-to-controller connection.A large backup path delay will significantly deteriorate the switch-to-controller delay.We define the average accumulated latency metric,which takes into account both the main path latency and backup path latency.We then propose a Reliability-aware Controller Placement Algorithm(RCP)for the controller placement problem.Finally,the performance of the proposed algorithm RCP is verified by simulation experiments.The experimental results show that the algorithm can optimize the delay on both the main and backup paths.(2)Research on multi-controller placement problem for improved both inter-controller and switch-to-controller latency.When multiple controllers are placed,communication is required between the controllers to maintain a unified understanding of the global network state.We consider both the maximum communication latency among the controllers and the average switch-to-controller latency,and formulate a novel multi-objective optimization problem.A Multi-objective Optimized Controller Placement algorithm(MOCP)combined with NSGA-II algorithm is proposed for the problem.The simulation results show that the proposed algorithm can achieve competitive performance in terms of the inter-controller and switch-to-controller latency.