On the efficiency of explicit multicast routing protocols

Abstract

Recently, many explicit multicast schemes have been proposed to serve a very large number of low to fairly moderate size multicast groups in a scalable manner. In this paper, we discuss the design trade-offs of these protocols both through simulation experiments and real-trees analysis. Our metrics for comparison are the header size and processing overheads. We also introduce a modification to ERM (explicit route multicast), called Bcast (branch cast), which removes some inefficiencies of ERM and reduces its size overhead. In Bcast, sender host generates the tree description code(s) and inserts it into header of multicast data packets. The code contains the IP addresses of all receivers and branching points of the tree. Bcast has a proactive bypassing mechanism which helps it to adjust the code size in response to inconvenient distribution of the receivers.