A Scalable Load Balancing Scheme for Software-Defined Datacenter Networks based on Fuzzy Logic

Abstract

In order to solve the problem that traditional load balancing technology lacks sufficient scalability and flexibility, we propose a dynamic and scalable load balancing scheme based on fuzzy logic for datacenter networks using the SDN architecture characteristics of control and forwarding separation, named LBSFL. The main parameters affecting the server load performance are analyzed and used as the input variables of fuzzy logic system to evaluate the server load, and then SDN controller schedules current network requests to achieve server load balancing. Additionally, we take into account the issue of server activation and deactivation under various traffic loads in this scheme. The experimental results show that the response time of the system can be increased by about 15% compared with LBBSRT, and it can also maintain the system load between the lower (30%) and normal (50%) boundaries. The scheme can greatly improve the load balance of servers under the premise of guaranteeing the network performance.