Imputato, Pasquale (2018) Network Traffic Control Design and Evaluation. [Tesi di dottorato]

Preview

Text phd-thesis.pdf Download (15MB) | Preview

Collection description

Recently, the term bufferbloat has been coined to indicate the uncontrolled growth of the network queueing time. A number of network traffic control strategies have been proposed to control network queueing delay. Active Queue Management (AQM) algorithms such as RED, CoDel and PIE have been proposed to drop packets before the network queues become full and to notify upper layers, e.g., transport protocols, about possible congestion status. Innovative packet schedulers such as FQ-CoDel, have been introduced to prioritize flows which do not build queues. Strategies to reduce device buffering, e.g., BQL, have been proposed to increase the effectiveness of packet schedulers. Network experimentation through simulators such as ns-3, one of the most used network simulators, allows the study of bufferbloat and to evaluate solutions in a controlled environment. In this work, we aligned the ns-3 queueing system to the Linux one, one of the most used networking stacks. We introduced in ns-3 a traffic control module modelled after the Linux one. Our design allowed the introduction in ns-3 of schedulers such as FQ-CoDel and of algorithms to dynamically size the buffers such as BQL. Also, we devised a new emulation methodology to overcome some limitations and increase the emulation fidelity. Then, by using the new emulation methodology, we validated the traffic control module with its AQM algorithms (RED, CoDel, FQ-CoDel and PIE). Our experiments prove the high fidelity of network emulation and the high accuracy of the traffic control module and AQM algorithms. Then, we show two proposals of design and evaluation of traffic control strategies by using ns-3. Firstly, we designed and evaluated a traffic control layer for the backlog management in 3GPP stacks. The approach improves significantly the flows performance in LTE networks. Secondly, we highlighted possible design flaws in rate based AQM algorithms and proposed an alternative flow control approach. The approach allows the improvement of the effectiveness of AQM algorithms. Our work will allow researchers to design and evaluate in a more accurate manner traffic control strategies through ns-3 based simulation and emulation and to evaluate the accuracy of other modules implemented in ns-3.

Downloads

Actions (login required)

View Item