Test Framework for Quality of Service Mechanisms for Multilayered Distributed Topology in a Software-Defined Networking Environment

• Main Author: Regencia, Josiah Eleazar
• Format: text
• Published: Archīum Ateneo 2021
• Subjects: n/a
• Online Access: https://archium.ateneo.edu/theses-dissertations/521
• Institution: Ateneo De Manila University

Software-Defined Networking (SDN) is a relatively new technology that provides dynamic, flexible and scalable control and management for networks. It is important to study the effects of applying Quality of Service (QoS) algo- rithms closer to the edge and how topologies affect their performance. Hence, a reusable framework for testing QoS algorithms and various network topologies in order to aid future research for QoS in the context of SDN is introduced in this study which is then used to explore how layers in the network topology af- fect the application of QoS algorithms. This is in order to replicate real-world networking scenarios where packets travel through layers of networking equip- ment. Simulations were performed using synthetic HTTP, streaming, and ICMP Ping traffic. Generally, results showed that Leaf-enforced QoS had comparable or better results to Core-enforced QoS. Leaf-enforced QoS had an average ad- vantage of 1% for both web throughput and streaming throughput. For Raw data throughput, Leaf-enforced QoS showed advantages in 2 layers and 5 layers test cases by 0.25% and 1.25%, respectively. Lastly, for latency ICMP Ping pack- ets, 2 layers up to 5 layers test cases showed a Leaf-enforced QoS advantage by an average of 1.9%. Increasing network traffic flows through the use of the iperf3 tool emphasized the performance advantage of the distributed nature of SDN over the traditional centralized networking architecture. For all test cases, vi Leaf-enforced QoS algorithms showed significant advantage in latency and ap- proximately 100% performance advantage for raw IP data throughput. This is an advantage considering that network traffic continues to rapidly increase and as networking infrastructure continue to become more and more complex.