Building a QoS Testing Framework for Simulating Real-World Network Topologies in a Software-defined Networking Environment

Software-defined networking is an emerging technology for implementing Quality of Service (QoS) for its logically centralized control and decoupled control and forwarding planes. In this paper, we extend a previous framework that simulated a fat-Tree topology software defined network under QoS provi...

Full description

Saved in:
Bibliographic Details
Main Authors: Yoo, Hyeong Seon, Yu, William Emmanuel S
Format: text
Published: Archīum Ateneo 2022
Subjects:
QoS
Online Access:https://archium.ateneo.edu/discs-faculty-pubs/358
https://doi.org/10.1109/ICEET56468.2022.10007108
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Ateneo De Manila University
Description
Summary:Software-defined networking is an emerging technology for implementing Quality of Service (QoS) for its logically centralized control and decoupled control and forwarding planes. In this paper, we extend a previous framework that simulated a fat-Tree topology software defined network under QoS provisioning techniques to support different topologies, including data from the Internet Topology Zoo. We compare round-Trip times between each server and client, and transfer rates of HTTP traffic in the tree, mesh, and topology provided by Kreonet and Uni-C under a distributed class-based queuing system. We observe that while round-Trip times are not significantly different, transfer rate of HTTP traffic as generated by ApacheBenchmark in Kreonet decreased 57.1109% from the fat-Tree topology, and 49.5328% from the mesh topology. Uni-C topology took 114.9% more time for the switches in the network to have installed all their flows than the Kreonet topology, the next slowest topology. Uni-C topology also failed to finish the HTTP benchmarks due to its more complex topology in a resource-limited machine. These differences demonstrate that the complexity of real world network environments (i.e. route loops, layering, convergence times) affect the network behavior, which is key to a good simulator. Therefore, the modified framework is able to support and properly simulate real-world network topologies.