Exploring the effects of monitoring and flow-rule timeout durations on load balancing in software defined networks

As more complex and distributed applications come into play in today´s Internet architecture, the need to have the adequate network architecture to support them with greater speed at lower costs has gained importance. Evolution in the nature of the applications have started to put significant stress on the networks and data centers which have to accommodate the demands of the highly distributed applications. Even though dedicated load balancers for each server can be used, this method is not cost effective and it is vulnerable to congestion and failure. In order to circumvent this problem, dynamic load balancing schemes, which adapt themselves to real-time network traffic, should be adopted. Frequent monitoring of the network traffic allows the load to be distributed fairly over the resources, but since it increases the communication overhead in the network, it deteriorates the system performance in terms of throughput and latency. In this work, our goal is to investigate how fairness and network performance are affected by fine tuning the monitoring and flow rule timeout parameters of an OpenFlow-based dynamic load balancer for data centers which uses real-time network traffic to make traffic-adaptive routing decisions.