Abstract

As group applications have become more prevalent, efficient network utilization becomes a growing concern. Multicast transmission may use network bandwidth more efficiently than multiple point-to-point messages, however, creating optimal multicast trees is prohibitively expensive. For this reason, heuristic methods are generally employed. These heuristics are often based on a Steiner tree approach, which is known to produce multicast trees that achieve an efficient use of network resources. Many such algorithms, both centralized and distributed, have been proposed to generate ‘good’ multicast trees. Even these heuristics typically have significant execution times, however, so changes to the initial group of multicast participants during generation of the tree is likely. Furthermore, periodic rebuilding of multicast trees or sub-trees has been proposed to improve the efficiency of these trees as the group membership evolves. Changes in group membership are also possible during this rebuilding process. Existing algorithms, however, either do not support changes to the multicast group during building of the tree or they impose unrealistic restrictions, such as no overlapping modifications or regeneration of the tree after every change. These restrictions prevent the use of such algorithms in many situations, e.g.; networks with mobile hosts. To remedy this, we propose an efficient distributed algorithm that supports dynamic changes to the multicast group during tree building and allows concurrent join/leave operations. In this paper, we present the algorithm, a proof of correctness, and detailed simulation results.

Author Keywords: Multicast protocols; Dynamic multicast groups; Distributed Steiner algorithms; Mobile networks

Article Outline

1. Introduction

2. Background

2.1. Related work

2.2. Problem statement

2.3. System model

3. The algorithm

3.1. Overview

3.2. Dynamic algorithm

3.2.1. Addition requests

3.2.2. Deletion requests

3.2.3. Combining fragment information

3.3. Termination condition and tree refinement

4. Correctness argument

5. Simulations

5.1. Simulation results

6. Conclusion and future work

Acknowledgements

appendix a. Algorithm pseudocode

References

Corresponding author. Tel.: +1-313-577-5474; fax: +1-313-577-6868