Fault Management in a Multicast Routing Environment: Kernel Based Tree Protocol, a Case Study
Mohamed Dâfir Ech-Cherif El Kettani, Younes Souissi · 2000
The conception of multicast routing protocols relies generally on the optimisation of Quality of Service parameters. Unfortunately, we rarely find propositions that consider as a criterion to enhance QoS associated to the group: We mean by fault tolerance the ability to manage failures of nodes and links of the distribution tree, in the network. A tolerant approach at the routing level will allow more reliability at the network and service levels. Currently, in a multicast routing environment, when a failure occurs, the solution consists in destroying the distribution tree, and starting from the beginning to construct again the whole distribution tree, which is a costly solution. We propose a solution avoiding that, by giving tolerant extensions to multicast routing protocols. Our proposition is based on the resolution of the consensus problem. Consensus problems were studied by Chandra and Toueg: We propose an inter domain version of the algorithm. We put the emphasis on scalability issues, due to the hierarchical structure of the network, such as the Internet. We study the case of KBT protocol, a QoS-sensitive protocol, adapted to inter domain multicast routing.