Abstract: The efficient utilization of bandwidth in campus networks is a major traffic engineering issue. It requires a complete knowledge of the underlying physical network architecture as well a means to automate or reactively and proactively program the network. The static nature of traditional network creates a hurdle that must be overcome to achieve the above. The Software Defined Network architecture proposes a novel way to automate, program and dynamically configure computer networks. This work uses the VMware virtualization software and the GNS3 network emulator to emulate a Software Defined-based campus network. A data plane made up software-based replicas of network devices is designed and configured to connect to a controller software. A network application scheme is implemented by leveraging the Hierarchical Token Bucket Queuing Discipline which automatically programs bandwidth allocation at the data plane through the controller based on traffic demands. The functionality of the architecture is tested by carrying out a number parallel-connections to simulate changing traffic patterns. This is done using the Iperf Application. The results show the conversion of a traditional campus network into a Software Defined-based campus network. It also depicts the complete emulation of the entire Software Defined-based campus network. At the data plane of the emulated network, devices are able to forward packets to one another with the most active port forwarding about 9,000 packets. The controller obtains a global view of all 11-network devices in the emulated network. The latency between the controller and the software defined switches at the data plane ranges between 50 and 62.5 milliseconds. The throughput between the controller and the software defined switches at the plane ranges between 2 and 9 Mbps. Application Plane to Control Plane communication in the emulated network is executed in an average of 30 milliseconds and bandwidth utilization occurs in a minimum of 11seconds and peaks at 27.5 seconds. It however becomes steady at 17 seconds as traffic patterns vary.
Keywords: SDN, virtualization, GNS3, Hierarchical Token Bucket, automation.
|
DOI:
10.17148/IJARCCE.2020.9617
