Research On OpenFlow-based Load Balancing Of SDN

In recent years,the development of cloud-based technologies has led to more complicated services.Therefore,the management and effective utilization of the underlying network resources are the future directions for development.Software Defined Network(SDN),with control plane and data plane decoupling,flexible programmability,and more supremely accurate flow control,combines OpenFlow that is a southbound interface protocol to collect and send network resource information to the controller.Load balancing technology is mainly used in distributed systems to improve cluster performance,optimize resource allocation,shorten service response time and increase system throughput.Therefore,this paper designs a load balancing scheme based on OpenFlow protocol to solve the load balancing problem of complex data center network.Firstly,this paper study SDN and load balancing technology,and design a dynamic load balancing scheme.The designed system is divided into basic network module,server load balancing module and link load balancing module.Dynamic server load balancing use the Simple Network Management Protocol(SNMP)to collect real-time parameters of a server cluster and evaluate the weights of each server based on the Analytic Hierarchy Process(AHP).Link load balancing obtains the number of flow tables maintained by the switch through OpenFlow messages and the port status to evaluate each link.Finally,K-1 redundant paths are calculated according to the shortest K path algorithm and a reasonable path adjustment threshold is set to adjust the path.This paper simulates under the Mininet and Floodlight controllers,verifies its feasibility through a reasonable test scheme and compares it with the static load balancing scheme.The dynamic load balancing is 10% ~ 20% faster than the static load balancing in the server response time,and the average throughput is about 3% higher,and the CPU load ratio fluctuates less.The research results show that the load balancing scheme designed in this paper can achieve dynamic load balancing and performance better than static load balancing.