After the birth of the Internet in 1970s, the Internet has grown exponentially because it was not only used for academic and military purposes but also as commercial demand and the development of WWW (World Wide Web) in 1990s. While in the process of deploying the Internet, multicast technology was introduced, which can deliver not multiple copies of message but even a single copy of message to all members of its destination host group. IP Multicast technology still has several problems to solve even though much elaboration for research and deployment has been done since 1992. The best environment for receivers is to have network fault free environment. If it is not possible, the next best thing is fast detection and accurate fault isolation when a fault has occurred.
FAULT isolation is defined as a task of locating the on-tree router(s) or on-tree link(s) which is the origin of a fault. We use the term fault when it results in a change to at least one receiver’s path from the source. A fault may result from the network problems such as router crash, link failure, and significant packet loss, etc. Fault isolation is essential for network operators or system administrators who are responsible for supervision of their corresponding network. However, isolating a fault in multicast tree is difficult because of the multicast characteristics such as dynamic nature of group membership, anonymity of members, and use of User Datagram Protocol (UDP) that provides no feedback.
Several models have been suggested to diagnose a fault in multicast trees, but they are generally limited to single-source sessions. In practice, there are many multi-source multicast sessions, such as large-scale replicated database, large-scale multimedia conferencing, and Web cache protocols. If we apply the single-source models to a multi-source multicast session, the number of control messages and time required to isolate a fault increases exponentially.
In this paper, we propose an efficien...