Qos-based multicast routing for real-time communications

Abstract

This thesis studies the distributed multicast routing algorithms that provide quality of service (QoS) guarantees in networks. It focuses on the delay constrained minimum cost tree (or constrained Steiner tree) problem; that is, it constructs a multicast tree that not only meets the constrained end-to-end delay requirements but also is the minimum cost multicast tree. This problem is known to be NP-complete. The existing distributed multicast routing algorithms that are based on heuristics do not consider a network environment where node failures occur. As a result, these algorithms will fail to complete the construction of a multicast tree when node failures occur during the construction period. Moreover, they will fail to maintain the constructed multicast tree if node failures occur after the construction period and during the on-going multicast session. In both cases, these algorithms will have to be restarted. We propose two new distributed delay constrained multicast routing algorithms that provide QoS guarantees even in networks where node failures occur. One is shortest path (SP) based and the other is minimum spanning tree (MST) based. They are capable of constructing a delay constrained multicast tree when node failures occur during the tree construction period and recovering from any node failure in a multicast tree during the on-going multicast session without interrupting the running traffic on the unaffected portion of the tree. The proposed SP-based or MST-based algorithms perform the failure recovery efficiently, which give better performance in terms of the number of exchanged messages and the convergence time than the existing distributed SP-based or MST-based delay constrained multicast routing algorithms in a network where node failures occur, respectively. The existing distributed delay constrained multicast routing algorithms can be classified either as an MST-based heuristic or as a SP-based heuristic. Two representative algorithms i.e., DKPP and DSPH are MST and SP based heuristics, respectively. The MST based algorithms like DKPP run with a high message and time complexity of O(n3). (Abstract shortened by UMI.)