Evaluation of PIM and CBT Multicast Protocols on Fault-Tolerance

Abstract

In high-speed communication networks, a failure in a multicast node or on a link is often possible. It is especially critical to understand how these failures affect the transmission of real-time applications in convergent data networks. This paper presents the fault-tolerance comparison of multicasting protocols concentrating on Protocol Independent Multicast (PIM) andCore-Based Tree (CBT) Protocols. We investigate the situations in which a copying node fails its multicasting function. A simulation model is developed such a way that faults are dynamically created on nodes and links “for a predetermined period of time.” This situation allows us to conceive how stable current multicast protocols are. The number of destination nodes receiving no copies due to faults is compared. The model developed in this paper repeats the simulation for several different size networks from the smallest of a source and two multicast destinations up to a network with a source and 580 different destination nodes. It will be shown that Protocols reactions to the network size are unexpectedly different. This paper presents results of a model of wide area multicast network where faults are introduced dynamically and randomly. Performance metrics in a wide area multicast network are measured and analyzed. The analysis reveals that how PIM protocol outperforms CBT in fault-tolerance for large-size networks in spite of the superiority of CBT in multicasting data under non-failing nodes presented in previous research.