An Efficient Routing Algorithm to Improve the Performance of Floodlight SDN Controllers

Abstract

With the advancement of high-tech devices and the growing demand for large data processing, networking systems have become unresponsive to user demands. Due to the high traffic on the network system, the capacity of network technologies such as wired, wireless, and cellular networks have been dramatically increased. The traffic nature of today’s network system is highly complex, making conventional networks incapable of handling it. These traditional network characteristics could not be adapted to changing demands. As a result, managing the various networking devices is extremely difficult because of the inflexibility of increasing network size and the reliance on the specific vendor’s software. Software-defined networking was created by decoupling network control from data forwarding devices. SDN aimed to create a networking paradigm that responds quickly to changing network requirements. The SDN controller uses the OpenFlow protocol, which handles rules for traffic at the switch. The Floodlight controller uses Dijkstra’s as a shortest path finding algorithm. Dijkstra’s algorithm searches for the shortest path without prior information. This makes Dijkstra’s algorithm waste valuable resources on useless nodes. In this thesis, the multi-heuristic shortest path algorithm, which removes unnecessary node checking, was proposed to optimize the performance of the Floodlight SDN controller. This study was implemented and tested on the network that the Mininet network emulator created. The network consists of virtual switches and hosts that the Floodlight controller manages. After network setup, the performance evaluation was performed using the CBench tool, which tests for throughput and latency metrics. Also, iperf is used to test packet loss. The proposed multi-heuristic shortest path algorithm can increase the throughput by an average of 22.27%, reduce an average latency by 5.63%, reduce response time by an average of 3.34%, and also decrease packet loss of 17.09% when compared with Floodlight controller with Dijkstra’s algorithm