Machine Intelligence Research Labs (MIR Labs)
This paper provides analytical tools to assess the improvements of existing, modified or entirely new multicast protocols at an early stage of design. In particular, the authors investigate the general properties of Sparse Mode (SM) and Dense Mode (DM) protocols and their variants. Specifically, they quantify the performance of these protocols in a ring topology network using Control Bandwidth Overhead (CBO) as cost metric. Their cost models are designed using combinatorial and restricted partitioning techniques. They compared their results with exhaustive enumeration technique and found that when p>0.4 (p is the probability that a router is part of a multicast group), DM with State Refresh (DMSR) mechanism appeared relatively superior to SM and DM with Flood and Prune (DMFP) mechanism.