New center location algorithms for shared multicast trees

Abstract

Multicast routing algorithms such as PIM, CBT, BGMP use shared multicast routing trees and the location of the multicast tree has great impact on the tree cost and the packet delay. In this paper we propose new center location algorithms and a new center relocation algorithm and analyze their performance through simulation studies. The proposed center location algorithms try to find the geographic center of multicast members considering not only multicast group members but also a few non-member nodes which are carefully chosen. Simulation results show that the proposed algorithms find the better center than existing algorithms in terms of tree cost and packet delay. After many members have joined and/or left the group, the previously chosen center may not be a proper place any more and, therefore, we need to find a new center and build a new tree around this new center. We propose a new center relocation algorithm that determines the moment when the new tree should be built around the new center. The algorithm is based on measured packet delays as well as the parameter indicating how much the group has changed. It not only avoids unnecessary center relocation processes but also prevents the cost and worst packet delay of the tree from significantly deviating from the optimal values. © Springer-Verlag Berlin Heidelberg 2002.