Research On Load Balancing Algorithm Of SDN Multi-Controller

As a new type of network architecture, SDN separate the control function from the traditional equipment, so as to realize the centralized control of the control plane. However, facing the increasing scale of modern network, the centralized control plane would become the system bottleneck. To solve of these problems, multi-controller distributed control plane solution has emerged. In the design and implementation of a multi-controller control plane, the static configuration between the switch and the controller can not adapt to the dynamic changes of the network flow, which can result in the load imbalance between controllers, and makes the controller become the bottleneck of the system. How to dynamically configure the mapping between the controller and the switch, maximize the performance of each controller, reduce the response delay of the controller and improve the system throughput, is the core issue of load balancing.According to the above problem, considering the real-time load and its processing capacity of the controller, this paper put forword two load balancing algorithm, a dynamic adjustment load balancing algorithm PALB which based on the system equilibrium and a load balancing algorithm IASM which based on auction and load characteristics. The PALB algorithm mainly aimed at the situation of overload of the signal controller,and based on principles of the overall system load, which ensure the high performance of the controller to allocate more load, so as to solve the load imbalance problem among the multi-controller. While facing multiple controllers overload at the same time, PALB is relatively simple, and easy to cause low migration efficiency. The IASM algorithm will auction the remaining rescorces of the controller, and make full use of the resources of each controller, so as to realize the optimal configuration of the controller and switch, and in each migration it can migrate more than one switch, so greatly reducing the network overhead.Experimental results show that PALB and IASM all can reach a certain amount of load balancing effect, and make full use of the resource of the control plane, in the case of multiple controllers overload, the IASM algorithm with low complexity and fast convergence rate, can quickly achieve dynamic deployment of the controllers and switchs.