Load Distribution of Software-Defined Networking Based on Controller Performance

Abstract

Software-Defined Networking (SDN) enabled by OpenFlow is a new paradigm of the network that is managed by a single centralized controller. In a large-scale network, distributed controller mechanism using multiple controllers is introduced to solve a problem of single point of failure. This mechanism requires load balancing to balance the loads of the controllers. The current load controller methods still have some issues. This work proposes a load balance method called Load Distribution mechanism and an algorithm based On controller Performance (LDOP). The LDOP can reduce loads of the overloaded controller by migrating switches to other controllers. Moreover, LDOP can reduce the impact when many controllers are overloaded. In this work, we evaluate the performance of LDOP by simulation. The experimental results show that LDOP mechanism can reduce loads of overloaded controller and does not make other controllers become overloaded. In case of many controllers are overloaded, it also reduces the impact of this. In addition, textbfLDOP also reduces the number of messages that controllers use in load balancing compared to other protocols.