Performance Characteristics of Selected Network Topology in a Software-Defined Networking QoS Testing Framework

Quality of Service in SDN/OpenFlow can be provided by making use of queues and their management techniques. This allows for better and more fine-grained control of network traffic. In this study, we test a Class-based Queuing (CBQ) algorithm for QoS provision that classifies network traffic according to the application (e.g., video traffic or HTTP traffic) over a selection of topology forms from the Internet Topology Zoo (ITZ). To accomplish this, we utilize a Mininet-based QoS Testing framework that allows for testing of various QoS algorithms. To select the networks to be tested, we use K-means clustering over the ITZ dataset according to network size for random sampling within each cluster. We show that under CBQ, HTTP stress test performance was better in clusters with smaller network size, while VLC video performance remained similar throughout clusters. KPI analysis also shows that clients closer to servers exhibited better performance than those farther away, while VLC video performance remained consistent throughout the network. The results show that the framework is able to satisfy QoS requirements while showing performance differences across various simulated network topologies.