Controller Adaptive Load Balancing for SDN Networks

Software Defined Network (SDN) is a novel network architecture which decouples the control plane from the data plane. One drawback of SDNs is the inability to survive when facing failures, in particular in large scale data center networks. Since SDN is programmable, flexible dynamic mechanisms could be designed to achieve fault tolerance. The paper studies the single master multiple slave controllers architecture to study fault tolerance. The paper proposes a Controller Adaptive Load Balancing (CALB) model for SDNs. CALB dynamically adapt the load of the master controller based on the current load at each slave controller. Mininet simulation tool is for the experimentation phase. Floodlight is used to implement the controllers. The key contribution of the paper is proposing a load balancing algorithm that the CALB is using to distribute the load among slave controllers based on a desired ratio. Experiments results were conducted using CALB in two scenarios: (1) Master controller is distributing switches among two slave controllers, and (2) master controller is distributing switches among three slave controllers. The paper uses throughput and response time for performance measurement. Results are promising as the performance increased 12% in terms of the throughput and 9% in terms of the response time when compared to a reference model Enhanced Controller Fault Tolerant fault tolerance technique (ECFT).