Linkcast: Fast and Scalable Multicast Routing Protocol

Abstract

Abstract. The current multicast routing protocols require state maintenance in the on-tree routers in order to forward multicast packet properly. Therefore, the routers memory may be easily saturated when there are very !arge nurober of low to moderate size multicast groups. In contrast, the explicit multicast protocols offer a stateless design using header space of multicast data packets. In this paper, we introduce new stateless scheme called Linkcast that efficiently eliminates processing overhead of explicit multicast protocols like Xcast. The Linkcast represents the multicast tree by encoding its counterpart links. Simulation results show that Linkcast completely eradicates the required unicast Iookups in explicit multicast protocols with less header size overhead.

Optimal False-Positive-Free Bloom Filter Design for Scalable Multicast Forwarding

“…The next theorem presents the main result of Section IV-B Theorem 1: The expected false-positive filter length can be approximated by the following formula: (13) where is the inverse function of the left branch 3 of . The statement of the theorem immediately follows from Lemma 1 and the symbolic solution of the equation with respect to (14) The approximation is worth rephrasing as (15) where is a very slowly increasing function of ( is approximately proportional to 2 From a practical point of view, the "horizontal" relative difference has higher significance than the "vertical" one. That is, for given prescribed tail probability , the smallest and for which , are to be found and compared as a "horizontal" relative difference .…”

“…This flat type of stateless multicast routing protocol (like Xcast, Xcast+ [13]) undoubtedly has the drawback that every intermediate node along the multicast tree should process the header, even in the case the node is not a branching point of the tree. To avoid mandatory packet header processing, tree encoding schemes have been proposed to integrate into multicast protocols, like in ERM and Linkcast [14], [15]. In these methods, the entire tree structure is encoded into the headers.…”

“…Based on thorough numerical assessment of the effect of this rounding procedure, we can state that all the numerically computable formulas and their analytical approximations for the required test range and the expected filter length change negligibly. Moreover, it can also be demonstrated that it is enough to use a single integer number of hash functions, and a suitable choice for this is (19) due to the approximation in (15). An illustrative example for the robustness of the filter length calculation against the rounding effect is as follows: The expected minimum filter length is 161.2 for , based on (9), and its approximation is 159.1 based on (15) (using real-valued numbers of hash functions).…”

Encodings of Multicast Trees

“…In Linkcast [11], a multicast tree is encoded using the link indexes of links in the multicast tree. Each forwarding node reads its corresponding link indexes and forwards the multicast packet.…”

“…Indeed, there are two proposals which encode multicast trees non optimally [11,13]. In this paper, we consider multicast trees which could occur at network or overlay level and show the correspondence of these trees to theoretical tree data structures.…”