LBFT: Load Balancing and Fault Tolerance in distributed controllers

Abstract

Software Defined Networking (SDN) at first was introduced with only one single controller. However, this causes serious problems such as scalability, availability and single point of failure. To overcome these problems, researchers have introduced distributed controllers as an initial solution. Nevertheless, we still have to deal with load distribution and fault issues among multiple controllers. In this paper, we propose a load balancing and Fault tolerance (LBFT) method for distributed controllers by using group migration of switches. The proposed method detects the failure of the master controller and try to migrate its connected switches to slave controllers. Meanwhile, it also tries to balance the loads among controllers. Our experimental results show that our approach not only can balance the loads among controllers, but also can re-optimize the overall throughput in short amount of time after failure. Furthermore, the failover process takes around 50 ms. Our method prolongs a little end to end RTT, and also there is no packet loss. Our proposed method can be implemented and run as a module on top of any controller.