Performance Evaluation of Algorithms of Dynamic Rendezvous Point Relocation

Abstract

The Protocol Independent Multicast - Sparse Mode (PIM-SM) uses one center (referred here as the Rendezvous Point “RP”) for all sources in a multicast group. PIM-SM distributes the multicast traffic of a source through a so-called shared distribution tree, whose root is at a predefined core called Rendezvous Point (RP). It also builds source-specific trees to the sources whose data rates exceed a defined threshold. In the literature, several investigations are done to improve and provide an efficient mechanism for the dynamic relocation of the RP depending on the sources or the members of the multicast group. In this paper, we extend the investigation of three search algorithms used to find the optimal RP position. To evaluate the performance of these algorithms, Estimated Tree Cost (ETC) and our improvement Enhanced Estimated Tree Cost (EETC), are used. The reason behind our choice these two methods is a comparative investigation of the RP-selection methods proposed in the literature. From the comparison we can see that ETC finds the most optimal position of the rendezvous point. The Hill-Climbing algorithm and the standard PIM-SM protocol with static RP-selection are used as a reference for comparison. Our algorithms result in a lower network load compared to RP-selection algorithm. However, they need additional control messages.