Local dynamic QoS control for large real-time multimedia sessions

In traditional sender-based rate adaptations, when network congestion occurs, the sender decreases its transmission rate. This rate can decrease to the level that suits receivers in the more congested regions or who are connected through slow access links. As a result, the quality deteriorates for many members, especially in large sessions, who do not have these problems and the bulk of members are disturbed. Because or the heterogeneous capabilities and the conflicting bandwidth requirements of different receivers in the same multicast group, this global adaptation becomes inefficient. We propose a scalable scheme for real-time multimedia adaptive applications. Our scheme has a hierarchical structure which organizes members dynamically in a hierarchy. Members are grouped into local regions. Each region has an agent (AG). Members are classified into two classes according to their network connection, whether they are connected through high speed or low speed connections. If congestion occurs in the local region, its agent acts as a relay sender adapting locally to congestion in its region, rather than having the sender adapt globally for the whole session. In addition, local adaptations may be performed only on one class of local members affected by the congestion.