A Comparison of Queue Management Algorithms for Intra-Flow Loss Control

|Multimedia ows are inherently inhomogenous, i.e. packets within a ow are of di erent importance for actual user perception. When transmitting such ows with real-time constraints in networks like the Internet which o er no reliability of transmission, some packet loss is inevitable. The perceptual impact of these losses is then ampli ed by the arbitrary distribution of packet losses within the ow which impairs the reconstruction of application data units (ADUs) at the receiver, as well as the performance of end-to-end loss recovery mechanisms. To control the loss distribution within a ow (\intraow" QoS) typically ltering higher-layer information within the network is proposed, which is both expensive in terms of resources, as well as undesirable with regard to network security. We compare two novel queue management algorithms which improve the intraow QoS without higher-layer ltering. The rst algorithm is called Di RED (Di erential Random Early Detection). It di erentiates between packets marked by the sender as either more or less eligible to be dropped in comparison to unmarked packets without keeping any perow state. On the contrary, the other algorithm called PLoP (Predictive Loss Pattern) operates without per-packet marking, yet with keeping partial perow state. We introduce simple metrics to describe the loss process of individual ows and present simulation results with voice as foreground tra c using the proposed methods in a multi-hop topology. We nd that both algorithms do not have a signi cant impact on the background tra c. For the given scenario algorithms using packet marking are found to be superior because for the foreground tra c a high probability for short bursts with potentially high perceptual impact can be traded against a higher probability for isolated losses as well as higher (but acceptable) probability for very