A performance comparison study of DCCP and a method with non-binary congestion metrics for streaming media rate control

This paper compares the performance of two algorithms for congestion control of streaming media. The two methods are Datagram Congestion Control Protocol (DCCP) over RED, and a solution based on an Active Queue Management (AQM) combined with explicit feedback of congestion level experienced at routers. DCCP relies on binary congestion metrics, either as packet dropping or ECN marking at AQM routers. In contrast, our proposed solution uses 32 bit congestion level metrics. Transmitted by ICMP Source Quench packets, this enables much faster and accurate response than the binary DCCP. The simulation tool ns-2 is used to compare the two methods transient and stationary behaviour, focusing on adaptation speed and accuracy, delay and delay jitter, and fairness. The results reveal that DCCP is inferior in almost all tests, and that the non-binary method proposed in this paper forms a sound network base to provide stable quality and controlled delay for rate adaptive streaming media.