Applying Wavelet-denoising in Congestion Control with Hybrid Internet Traffic
In the current Internet, congestion control is performed in a TCP/AQM feedback loop. At the end systems, Internet traffic is mainly controlled by TCP protocols, which apply an additive increasing and multiplicative decreasing (AIMD) congestion control scheme. To avoid oscillation of queue length at the bottleneck links and provide early congestion indication to the end systems, Active Queue Management (AQM) algorithms are applied at the bottleneck routers, e.g. RED is the de facto standard technique used in the routers. However, there is also unresponsive traffic, such as HTTP traffic, that does not responsive to AQM congestion control signals (i.e. packet loss/mark probability). In the real network, the control parameters of AQM algorithms are highly sensitive to the mix between long term TCP traffic and unresponsive traffic. Designing a robust TCP/AQM controller with the presence of a mixture of unresponsive and responsive traffic remains a challenging problem. Our design is the first to introduce a wavelet-based traffic de-noiser to remove the impact of unresponsive traffic on the AQM algorithms. By introducing a separate traffic de-noiser , we allow the AQM algorithm to remove the impact of unresponsive traffic while still being able to maintain the optimal parameters for the responsive traffic flows. This enables a new paradigm in the design of AQM controller with hybrid traffic in the Internet.
Fig. 1 Overview of the System Design
Presentations
Presentation on NetSys 295 in Fall'04
Presentation in IEEE ICCCN 2007
Publications
Y. Pan, W. Tsai, and T. Suda, "Applying Wavelet De-noising to Improve TCP Throughput in AQM queues with Existence of Unresponsive Traffic", in the Proc. of IEEE ICCCN 2007, Aug 2007. [PDF] Errata to ICCCN CD Proceeding
Source Codes
Matlab Simulink Simulator
NS-2 Simulator
Researchers: