Weighted-Fairness AIMD Congestion Control for Streaming Media Delivery

Abstract

Streaming media applications require acceptable levels of quality of service (QoS). Appropriate rate control mechanism is needed to make streaming media applications to adapt to dynamic network conditions. This paper discusses a window-based weighted-fairness AIMD (WF-AIMD) congestion control algorithm which allows competing flows to share a common bandwidth according to their weights. By making flows to reduce their congestion window sizes with respect to their weights upon detection of congestion indications, flows can be made to unevenly share a common network path. The queueing scheme adopted at the bottleneck link largely impact the bandwidth share of a flow. Following a fluid modeling approach, we derived the modeling of the sending rate of a flow that is controlled under a WF-AIMD algorithm. This modeling has been validated under the DropTail FIFO queueing scheme and the RED scheme. We found that our modeling works well for both the RED scheme and for the DropTail FIFO queueing scheme with a relatively small queue size limit. Our modeling fails when flows become synchronized. The synchronization usually happens when the bottleneck queue adopts a DropTail FIFO queueing scheme with a large queue size limit.