Incremental deployment of a router-assisted reliable multicast scheme


Recently, several schemes have been proposed that address the problem of scalable reliable multicast using router assistance. These include LMS, PGM, AIM and OTERS, among others. While these schemes (arguably) achieve significantly better performance than non-assisted schemes, a thorny issue remains: are these schemes incrementally deployable? More specifically, what is the performance impact of partial deployment to the application and the network? What is the best strategy to deploy such schemes? These questions, while very important, have not been addressed yet. In this paper we investigate the performance of Light-weight Multicast Services (LMS) under various deployment strategies. Our results are preliminary, mostly due to lack of established strategies for incremental deployment of such services. First, we propose a method to incrementally deploy LMS which has minimal impact on router processing and state. Then, we investigate the performance of LMS for various basic deployment strategies, including random, core, stub, and border router deployment, and for different topologies, including random and transit-stub topologies. We also examine node v.s. path deployment. Our evaluation is carried out using the ns simulator and topologies generated with GT-ITM. Our results show that there are significant differences among the various deployment strategies with random being the worst. Finally, we discuss our observations in the context of other router-assisted schemes and identify common issues.

Publication Title

Second International Workshop on Networked Group Communcation (NGC 2000)