Home // International Journal On Advances in Telecommunications, volume 8, numbers 3 and 4, 2015 // View article

ConEx Performance Evaluation and Application to Video Streaming

Authors:
Ali Sanhaji
Philippe Niger
Philippe Cadro
André-Luc Beylot

Keywords: ConEx; Performance; ECN; Congestion; Policing; YouTube; LEDBAT.

Abstract:
With Internet traffic ever increasing, network congestion should occur more and more frequently. During congestion periods, some users contribute more than others to the congestion in the network. It might be interesting for a network operator to differentiate between users proportionally to the congestion they induce, but the necessary information for this purpose is not available at the network layer, and is exchanged at the transport layer (e.g., Transmission Control Protocol (TCP) acks). This led the Internet Engineering Task Force (IETF) to design Congestion Exposure (ConEx), a new mechanism to expose to the network the amount of congestion a user is responsible for, allowing the network operator to improve the fairness between users. ConEx is designed to limit the added complexity, leveraging already existing mechanisms such as Random Early Detection (RED) and Explicit Congestion Notification (ECN), plus a number of modifications to the senders and receivers to be fully operational. Nonetheless, ConEx can also be deployed in a simplified mode of operation, relying only on loss information in DropTail queues to estimate congestion. The objective of this paper is to provide an in depth evaluation of ConEx mechanism. Firstly, we investigate how the setting of ConEx parameters (e.g., congestion policer) and the network configuration (e.g., router queuing, network delay, etc.,) impact the behavior of ConEx and influence its ability to improve fairness between users. Secondly, we compare the level of performance, in terms of fairness improvement, provided by different variants of ConEx of increasing complexity, i.e., from a simple implementation with modifications limited to the sender to a “full” ConEx approach implementing all proposed features. We show that, despite a reduced accuracy in congestion estimation, a simple variant of ConEx is already able to provide a good fairness improvement between users. This is particularly interesting in the context of an initial deployment scenario, allowing an incremental deployment of ConEx. Thirdly, we investigate and discuss the limitations and weaknesses presented by ConEx with regard to short-lived flows. Finally, based on a YouTube traffic model, we illustrate how ConEx can help to enhance the Quality of Experience (QoE) of video streaming users during congestion periods, significantly reducing the number and duration of stalling events.

Pages: 202 to 214

Copyright: Copyright (c) to authors, 2015. Used with permission.

Publication date: December 30, 2015

Published in: journal

ISSN: 1942-2601