Evaluation of Active Queue Management (AQM) Models in Low Latency Networks
Abstract: Low latency networks require the modification of the actual queuing management in order to avoid large queuing delay. Nowadays, TCP’s congestion control maximizes the throughput of the link providing benefits to large flow packets. However, nodes’ buffers may get fully filled, which would...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2019 |
| País: | España |
| Institución: | Universidad del País Vasco |
| Repositorio: | Addi. Archivo Digital para la Docencia y la Investigación |
| OAI Identifier: | oai:addi.ehu.eus:10810/34832 |
| Acceso en línea: | http://hdl.handle.net/10810/34832 |
| Access Level: | acceso abierto |
| Palabra clave: | latency TCP AQM bufferbloat ABE FQ_CoDel latencia |
| Sumario: | Abstract: Low latency networks require the modification of the actual queuing management in order to avoid large queuing delay. Nowadays, TCP’s congestion control maximizes the throughput of the link providing benefits to large flow packets. However, nodes’ buffers may get fully filled, which would produce large time delays and packet dropping situations, named as bufferbloat problem. For actual time-sensitive applications demand, such as VoIP, online gaming or financial trading, these queueing times cause bad quality of service being directly noticed in user’s utilization. This work studies the different alternatives for active queue management (AQM) in the nodes links, optimizing the latency of the small flow packets and, therefore, providing better quality for low latency networks in congestion scenarios. AQM models are simulated in a dumbbell topology with ns3 software, which shows the diverse latency values (measured in RTT) according to network situations and the algorithm that has been installed. In detail, RED, CoDel, PIE, and FQ_CoDel algorithms are studied, plus the modification of the TCP sender’s congestion control with Alternative Backoff with ECN (ABE) algorithm. The simulations will display the best queueing times for the implementation that mixes FQ_CoDel with ABE, the one which maximizes the throughput reducing the latency of the packets. Thus, the modification of queueing management with FQ_CoDel and the implementation of ABE in the sender will solve the bufferbloat problem offering the required quality for low latency networks. |
|---|