Server-based Dynamic Load Balancing

In recent years, there is a rapid growth in the internet services which has increased incoming traffic in the network and also plays a vital role in increasing the size of the cluster server. The load balancing is a key technology to manage the traffic efficiently. In this paper, we propose an alternative load balancing architecture using OpenFlow switches which are connected to the controller. This gives high flexibility without any additional equipment. The system could measure network and server status and dynamically set weights of cluster servers according to the processing capability. In our approach, load balancer installs wildcard rules in the switches proactively based on the load of servers which is measured based on memory usage, CPU usage, and active connection counts. In our implementation, we used floodlight as OpenFlow controller and Mininet as network emulator to verify the soundness of the algorithm. The results show that our dynamic server load balancing algorithm is better than round robin and random load balancing algorithms.