Scalable session and traffic management for multiparty sessions

Abstract

The global network, e.g. Internet, provides the opportunities to the users to participate in interactive sessions. Powerful computers and high-speed networks made it possible for large size data such as video data and audio data interchanged among the participants. Internet multicast is a useful mechanism to disseminate the large size data to a lot of participants. We expect that a multiparty session can have more than hundreds of participants. A large number of participants results in two problems. First, interactive sessions require coordination to be successful and existing session control scheme will not work properly. Second, the network heterogeneity among the participants becomes more serious and it is difficult to disseminate large size data satisfying the various network conditions of the receivers. This paper suggests solutions to the problems. For the first problem, we suggest a hierarchical session model for the scalable session management. For the second problem, we suggest a network-level multicast traffic managment scheme. The hierarchical session model proposes that a session is divided into multiple subgroups when its size becomes larger than a predefined value. When the number of participants reaches the predefined value, a session divides into multiple groups and each group has its own group manager. The interactions among the participants are coordinated through group managers and a session manager. It reduces the number of objects managed by the a session manager. The traffic control scheme requires three basic models;application model, network model and traffic model. Application is able to encode data into hierarchically structured layers and to decode data delivered in forms of multiple layers. The first layer is essential to regenerate the original data. Time-to-Live(TTL) threshold is assigned to every out link and the value is modified according to the neighbor router``s request. Traffic controller located in a router measures inp...