Layered Multicast: A Study of Loss Event Rate Estimation in a Low Level of Statistical Multiplexing

Layered multicast protocol (LMP) uses TCP-equation model to estimate TCP-compatible rate. One of the most important parameter of TCP-equation model is loss event rate. It is acquired by estimating the number of packets between two lost events, which is determined by packet-drop pattern at the bottleneck link. In a low level of statistical multiplexing environment packet-drop pattern at the bottleneck link is determined by the queuing management and the behaviours of competing data flows. Since TCP is the dominant protocol of the Internet, the performance of LMP is greatly affected by the behaviours of TCP flows. Since TCP is an aggressive and bursty protocol, competing with TCP results in volatile estimated loss event rate for LMP. Smoothing techniques for loss even rate have been proposed to reduce the volatility of loss event rate. This paper presents a comparative study of loss event rate smoothing techniques for layered multicast protocol. The study has been conducted in a low level statistical multiplexing environment.