A Sharing Mechanism Based Controller Failure Recovery Approach In Multi-domain SDN

Software-Defined Networking(SDN)is a new architecture of network,which has attracted a lot of research and attention of industry,and it has been regarded as an important implementation of network virtualization.The core of SDN is dividing the computation modules from data forwarding modules of traditional network devices,and computation modules are centralized on a programmable controller,which computes the path of data forwarding and sends the corresponding flow table to the switches.Obviously,the controller plays an important role in SDN.When the scale of network becomes large,the controller cannot respond many requests in required time.Therefore,it is necessary to divide the network into many domains and deploy one controller to manage the network in each domain,which makes multi-domain SDN.Thus,how to ensure the reliability of the controllers in multi-domain SDN is a key problem.In order to ensure the reliability of the controllers in multi-domain SDN,it is necessary to backup the master controller of each domain in real time to prevent the network becoming paralyzed because of the master controller failure.About the problem of backuping the master controller in multi-domain SDN,it needs to consider two aspects,one is how to backup the network state information of master controller in each domain,and the other is how to deploy the backup controllers.For the first problem,the direct solution is making the master controller send the network state information to the backup controllers in real time.This solution can be used in SDN with a single controller.However,if the solution is used in multi-domain SDN,it may cause the extra communication between the master and backup controller.Additionally,the master controllers need to communicate with each other to get the whole network information.Therefore,this backup method may influence the performance of controller because the resource is occupied by communication.For the problem of how to deploy the backup controllers,the most common method is deploying one backup controller in each domain.But this may cause the waste of resource,since the probability of all controllers in each domain fail at the same time is very small.In other words,some backup controllers may be in idle state,and they only take up network resource without management of the network.To improve the reliability of controllers in multi-domain SDN,we need to design more effective method to backup the network information of the master controller in each domain,and deploy the backup controllers in multi-domain SDN reasonably.In this paper,we discuss the problem of backup the network information of the master controller and the backup controller placement problem in multi-domain SDN,the main research contents include:(1)We propose a controller failure recovery method based on the sharing data store,which is aimed to store the information of the network state based on the sharing data store in real time.This method can reduce the communication between master and backup controllers,and then improve the performance of the whole network.We first design the data structure to save the network information.Then we design the deployment of the sharing data store based on delay and hot-standby mechanism to improve the reliability of the sharing data store.Last,we propose the process of the controller failure recovery and repair based on this sharing data store.(2)We propose a backup controller placement based on sharing mechanism.It means many domains share a few backup controllers.In order to utilize the network resource effectively,it needs to deploy some backup controllers and place them on suitable situations.In this paper,we first compute the least number of needed backup controllers based on the probability of controller failure.Then we design three objective functions,the number of backup controllers,the delay between backup controllers and switches and the reliability of the network.Last we compute the solution of backup controller placement using PSO algorithm based on multiple objectives optimization.