In this research area we are looking for an approach towards group communication across ATM networks that provides support for heterogenous receivers and senders. Applications that request such services are conferencing tools, distance learning tools and distributed simulations. These applications are characterized by a group of participants that simultaneously act as recipients and sources of (real-time) information. In todays networks ATM mainly operate in the backbone area. However, no adequate support for heterogeneous group communication is provided.
The ATM technology enables the creation of bidirectional one-to-one and unidirectional one-to-many connections with specified QoS. However, ATM-UNI signaling does not support many-to-many service. This form of group communication service must be emulated. Additionally, the current ATM-UNI disallows changes to the QoS of an established connection, i.e. dynamic QoS is not supported.
Currently existing solutions to emulate IP-Multicast over ATM, like MARS and EARTH are not capable of supporting heterogenous QoS. They are only based either on a VC Mesh (n 1:n VC's) or MCS (MultiCast Server, n 1:1 connections to MCS and 1 1:n connection to receivers). Both schemes are not flexible enough to support heterogeneous group communication with dynamic QoS. A single MCS can become a bottleneck for a higher number of senders. The VC Mesh has a high resource consumption and produces a high signalling overhead due to group membership changes. Both schemes did not scale when the number of participants increases.
In our concept, VC mesh and multiple MCS are flexibly combined in order to provide support for heterogeneous group communication. Through the combination of these two schemes different QoS for heterogenous sources and destinations can be supported. In a MCS different data streams can be unified and delivered to several receivers. The VC Mesh technique can be applied, if the QoS requirements of sender and receiver are similar and the delay constraints are very high.
The main problem of the proposed approach is, to obtain a well suited configuration that serves the actual demands of a communication group. With a number of MCS and sources and recipients with different QoS demands the number of potential combinations increases exponentially. However, in contrast to VC mesh and MCS heterogeneous receivers can be accomodated according to their individual QoS requirements. The optimization problem is to provide a SVC distribution that minimizes network load and resource consumption. Especially dynamic changes of group membership enforce a flexible solution for this problem.