There are two basic multicast tree construction techniques, source specific trees and center specific trees. Source specific trees require finding a shortest path from the sender to each receiver, resulting in multiple minimal delay trees for a group. Center specific trees (or shared trees) use a distribution center(s) and construct a single multicast tree resulting in a low overhead method, sacrificing minimal end-to-end delay. Desirable features for a multicast data distribution service in a global network are:
The present Internet multicast protocol, Distance Vector Multicast Routing Protocol (DVMRP), specified in RFC 1075, is based on distance vector unicast routing using a source based data delivery method. The undesirable broadcast behavior of this method for groups whose members are distributed sparsely across a wide area is being addressed by the IDMR network working group of the IETF. Multicast Extensions to OSPF (MOSPF), based on the Open Shortest Path First (OSPF) unicast routing algorithm, is another Internet multicast protocol. Since it is not feasible to maintain a complete global link database at every node, MOSPF is also not a globally scalable solution. These two protocols are considered suitable for intra-domain multicast.
Two internet drafts being considered by the Internet Engineering Task Force (IETF) Inter-Domain Multicast Routing (IDMR) group are Protocol Independent Multicast (PIM) and Core Based Trees (CBT). CBT uses a center specific tree protocol for multicast tree construction. Sparse mode PIM combines both source specific and center specific trees for a single group to provide a scalable solution for groups that are not densely populated.
Sparse mode PIM and CBT use distribution centers, called rendezvous points (RPs) and cores, respectively. Placement of the distribution centers is presently described by both drafts as a task to be done administratively. The placement of the distribution centers has a direct effect on the quality of the multicast tree constructed. In turn, the quality of the tree affects the overhead of the network and the end-to end delay for the user. At present the network administrator has no tools to assist in the placement of these centers.
We present a tool which will assist the network administrator in the task of center selection. We describe the features, design and utility of a Multicast AnalysiS Tool (MAST), a menu driven tool with a graphical user interface that allows the user to analyze multicast tree quality in an internetwork topology. MAST supports a two level hierarchical internetwork. Clusters of local nodes are connected to reflect an intra and inter-domain routing architecture. The tool is made up of the following components:
We show simulation results on various topologies showing that, with intelligent, algorithmic placement of distribution centers, higher quality trees can be constructed. We graphically display the network traffic loads in inter- and intra-domain links. Finally, we demonstrate MAST on a realistic topology by modeling a portion of the Multicast Backbone (MBONE) to illustrate the utility of the tool.
With two independently developed center specific tree approaches proposed in the IETF, this work is expected to contribute to the decision process for choosing the standard. With a tool such as MAST, differences in the proposed protocols can be exposed or lack of such differences illuminated. Thus, MAST can be used for two diverse purposes, namely it can be used to gather data points for decision making in the standardization process or it can help with the proper placement of centers by network administrators in an operational network.