The Research And Design Of Load Balance For Controllers And Servers In SDN

Software-Defined Networking(SDN)is a new type of network architecture which has aroused extensive research and attention in the industry.As the demand of user’s increases,the scacle of network become larger,and SDN also become larger,therefore,multi-controller architecture is born for this demand.However,the multi-controller architecture has the problem of uneven load distribution,which the multi-node architecture must have.In addition,the proposal of SDN makes it necessary for us to further study the existing network problems to solve some problems,using the characteristic of SDN,including the load balancing of server clusters.combined with the features and requirements of software-defined network,this paper mainly studies the load balancing mechanism of multi-controller and server cluster in SDN,and puts forward the optimization schemes separately.The main contents of this paper include the following two aspects.(A)To resolve the problem of the uneven load distribution among controllers,the paper propose a load balancing scheme based on switches group.Based on the switch migration strategy,this scheme designs three modules:load monitoring,load balancing and switch migration.The load monitoring and switch migration module are deployed on the local controllers,respectively,which is responsible for monitoring and informing the load of controllers and switches,migrating switches.Load balancing is deployed on the global controllers,which is responsible for balancing the load of overloaded controllers.The switches group-target controllers group selection algorithm,optimal switches group-target controller selection algorithm and optimal target controllers selection algorithm are designed in this module,the algorithms select the optimal switches group and target controllers group under the conditions of management cost,migration cost and load balance from the global optimal.The algorithms not only optimize the network overhead and load balance,but also avoid the problem of frequently switches migration and the target controller become overloaded caused by multiple switches being migrated to the same controller.In addition,the idea of switches group and switches in a switch group are migrated at same time using multi-thread mechanism improve the time efficiency of the solution.(B)To solve the problem of uneven load distribution among server cluster,the paper proposed a server cluster load balancing scheme based on dynamic adaptive informing strategy.Based on the general process of the server cluster load balancing mechanism using the integrated load,this scheme designs the load pushing module and the load balancing module,which are respectively deployed on the servers and the controllers.We design an informing strategy in the load pushing module,and the strategy takes the server computing pressure,real-time and network overhead into account.This strategy adapts dynamic period to read the data,which can make server read the data timely,and server computing pressure not too big.In addition,the strategy improves the real-time performance of load in controller and reduces the network overhead.The multi-queue target server selection mode is designed in the load balancing module,which not only improves the server load balancing,but also optimizes the time of the server become overloaded caused by the big flow problem.