Delay analysis of application level multicast on content addressable networks

Two approaches have been proposed for application level multicast on content addressable networks (CANs), implemented through tree-based and flooding-based methods. In this paper, we present simple analytical models for studying these two approaches. Using these models, we study the performance achieved by multicast groups as a function of their size, the variability in delays between neighbor nodes and the dimensionality of the underlying CAN. We see that at lower dimensions and lower group sizes the CAN flooding scheme performs better than the tree-based scheme. However as the dimensionality or the CAN increases and the group sizes increase the tree-based scheme performs better. We also propose and study a scheme for reducing delays in the tree based multicast scheme.