Research On Placement Method Of SDN Controllers Based On Firefly Algorithm

With the explosive development of information technologies such as cloud computing,Internet of Things,and big data,higher requirements have been put forward for the network to process massive amounts of data,such as lower packet loss rates,millisecond delays,and more demanding security level.Traditional network architecture is difficult to meet these requirements,so a new type of network architecture was born: software defined network(SDN).SDN has the characteristics of separation of control and forwarding,centralized control,programmable,open interface,and virtualization,which solves the problems of traditional computer network hardware such as not being open and unable to be programmable.SDN separates the data forwarding function and control function coupled in the traditional network,and sets up independent control layer and data layer.The control layer controls the data forwarding of the underlying SDN switch by issuing a flow table.In SDN,the control layer occupies an absolutely important position.To ensure the robustness of the control layer,SDN usually adopts a logically centralized multi-controller placement method.The placement of multiple controllers needs to solve two key issues:(1)the location of the controller;(2)the mapping relationship between the controller and the switch.Studies have shown that the SDN multi-controller placement problem is an NP-hard problem,and the optimal solution cannot be obtained in a limited time.For this reason,based on the Firefly algorithm,this paper designs an SDN multi-controller placement algorithm,which can obtain a controller placement solution with low inter-node delay and balanced controller load in a limited time.The main contributions of this paper are as follows:(1)Mathematical model of multi-controller placement problem oriented to node delay and controller load.First,establish SDN as an undirected graph to define concepts such as controller nodes,switch nodes,controller-to-switch mapping relationship,controller load,etc.;secondly,for the delays between controllers and controllers and switches that cannot be the lowest at the same time The problem is to propose a comprehensive transmission delay to meet different controller placement requirements;finally,to propose a controller load and utilization index,so as to balance the controller load in the entire network.(2)Multi-controller placement method based on Firefly Algorithm.In order to solve the multi-controller placement model in a limited time,this paper designs a multi-controller placement method based on the firefly algorithm.Firstly,the two-dimensional coordinates of the controller to be placed are abstracted into the firefly vector in the algorithm;secondly,the integrated transmission delay is taken as the objective function of the algorithm;finally,the upper limit of the controller load utilization rate is taken as the constraint condition of the algorithm,thus Obtain a controller placement solution with low integrated delay and balanced controller load in a limited time.(3)Improvement of the Firefly Algorithm.To obtain a controller placement scheme with lower integrated transmission delay and more complex controllers in a limited time,this paper improves the Firefly algorithm from two aspects and uses it to solve the multicontroller placement problem.On the one hand,to address the disadvantages of the Firefly algorithm's poor global search ability and low post-convergence accuracy,iterative dynamic optical attraction,attraction factor,and random parameters are used;on the other hand,for the static initial parameter setting of the Firefly algorithm,particle swarm optimization is used.The algorithm optimizes the static parameters of the multi-controller placement algorithm based on the Firefly algorithm.The experiments of the real WAN topology in The Internet Topology Zoo show that compared with other controller placement algorithms,the controller placement scheme solved by the SDN controller placement algorithm proposed in this article has lower integrated delay and more balanced controller load.