Research On Load Balancing In Data Center Under Software Defined Network

Software Definition Network(SDN)is a revolutionary next-generation network architecture proposed by the Stanford University's Clean Slate Research Group.Unlike conventional networks,the SDN separates the control layer from the forwarding layer through a southbound interface protocol.As the control layer,SDN controller is responsible for the issuance of the flow table,mastering the topology of the network information and providing an open programming interface to the upper application.SDN switch is only responsible for forwarding network packets according to the flow table provided by the related SDN controller.SDN overcomes the shortcomings of the traditional network and promotes the network innovation.With the rise of mobile Internet and big data,the data center with traditional IP network architecture faces with a huge challenge.SDN,as an emerging network architecture,separates the control plane from the data plane.It has centralized control and programmable features,which is very suitable for data center.As the SDN itself is highly dynamic,how to improve the timeliness of its load balancing and to better solve the network congestion is essential.So in this paper,we study the load balancing technology of the data center in the software defined network.Firstly,this paper proposes a load balancing scheme for server cluster based on SDN and the improved ant colony algorithm.Then,a SDN link load balancing scheme based on particle swarm optimization algorithm is proposed.By using SDN controller's programmable features,both schemes can realize dynamic load balancing of data center by software.In the SDN server cluster load balancing scheme,we first use the SNMP protocol toperiodically obtain the load information of each server,and then use the improved ant colony algorithm to allocate the concurrent tasks to the corresponding server based on the acquired load information,ultimately minimizing the response time of the task,the cost of the service,and the load difference between the servers.Based on the flexibility of the SDN network,the scheme does not require additional facilities and has a strong scalability.By monitoring the load condition of SDN link in real time,SDN link load balancing scheme uses the improved particle swarm algorithm and SDN controller to realize the load balancing of each link of OpenFlow switching network,so as to make effective use of network resources and achieve the effect of link dynamic load balancing.