Research On Controller Placement Problem Based On Load Balance And Latency Constraint In SDN

The control plane is decoupled from the data plane in Software Defined Network(SDN),which can programmatically manage the network devices by centralized controllers,thus the network resources can be flexibly configured.However,with the development of SDN network,multiple-controller architecture is exploited to manage the network,which incurs the controller placement problem(CPP).That is,for a given network,how many controllers are needed,and where should they go?Some related researches show that well-designed strategies of controller placement can not only effectively improve network load balance and reliability,but also guarantee the network latency.In this thesis,we conduct research on CPP based on load balance and latency constraints in SDN,and main research contents are as follows.(1)Research on static multi-controller placement problem based on load balance and delay.The network requests will not be processed in time when the controller is overloaded.In order to deal with this problem,a static controller placement algorithm is proposed in this thesis.Firstly,the number of controllers,K,is determined based on controller's load capacities and switch requests.Secondly,a partition algorithm with approximate ratio of 2 is proposed to obtain K controller's management regions.Lastly,the location of controller in each area is elected according to the minimum sum of all switch-to-controller distances in relevant zone.As to compare with AL and WL algorithms,simulation results show that the proposed algorithm not only balances the controllers' load,with an approximation ratio of 2,but also guarantee the network latency.(2)Research on dynamic multi-controller placement based on load balance and delay.On the basis of the above research,the CPP is further studied with the consideration of network reliabilities and dynamic controller loads.And the controller placement is divided into three phases in this thesis.Phase one:a modified greedy algorithm is proposed to locate controllers in initial network.Phase two:based on the latency,load balance,reliability and connectivity,an algorithm is proposed to determine the management of each controller area.Phase three:when a controller is overloaded(the entire network is not overloaded),the controller's management area can be adjusted by an online algorithm to ensure the stable operation of the network.Compared with other methods,simulation results show that the proposed algorithms not only reduce the number of controllers but also ensure latency,load balancing and the reliability.