Abstract:Multicast multimedia distribution, despite the benefits it offers, features a number of unresolved issues, such as network and users heterogeneity. Layered stream organization has been considered as suitable to cope with these issues. In this paper we, initially, study proposed mechanisms for the multicast delivery of layered video streams, highlighting their principles and functionality. Most of them are suited for Best Effort networks, resulting in increased application complexity or sub-optimal performance. Based on the assumption that Quality of Service (QoS) will be probably deployed to the future Internet, we propose a simple, network-based architecture for the multicast delivery of layered video streams. The Differentiated Services (DifJServ) framework and content-aware packet marking are exploited in this architecture. We discuss, how the application defmes priority layers within the video stream, while DiffServ copes with layer-aware packet dropping. Each video stream is delivered to a single multicast group and no application-level complexity is required. During congestion in nodes of the multicast tree, packets corresponding to lower priority layers are dropped first, thus protecting the most valuable video stream entities. Protocol extensions for coping with user heterogeneity and the problem of congestion from undelivered packets are also proposed. Finally, a qualitative comparison of all the aforementioned schemes is attempted.