Application-Level Adaptive Congestion Detection and Control for VoIP

Abstract

For decades, researchers have worked extensively in the area of congestion control for packet-switched networks. Many proposed solutions take advantage of the congestion-control mechanism in transmission control protocol (TCP), and these approaches work well for networks that have heavy TCP traffic. However, these approaches are not universally effective - they fail completely for protocols that do not implement congestion-control mechanisms. In particular, these approaches do not work with the user datagram protocol (UDP). Realtime media-streaming technologies such as voice over Internet protocol (VoIP) and video conferencing use UDP and therefore do not respond well to existing congestion-avoidance techniques. We propose a new, adaptive, responsive, end-to-end technique to implement application-level congestion detection and control for real-time applications such as VoIP. Unlike existing methods, which rely on packet loss as a signal to reduce the transmission rate, our solution reacts to network congestion to prevent packet loss, thus improving the QoS of applications that employ our algorithm.