Routing Protocols

Multicast 1 is one of the best available service models for scalable, manyto-many delivery of data on the Internet. Multicast achieves scalability by letting sources send a packet only once, regardless of the number of receivers. In network-layer multicast, network elements (routers, switches, and so on) replicate packets at network branching points, delivering copies to all interested receivers. Replicating each packet as close to the receiver as possible improves efficiency. Multicast uses range from the delivery of conventional multimedia streams such as audio and video to emerging services such as desktop streaming, whiteboards, and collaborative distance learning. As the multicast infrastructure has matured, several limitations have come to light. Although the infrastructure has recently become more robust,2 weaknesses in the underlying protocols have historically resulted in poor connectivity and significant routing instabilities. Although the severity of multicast routing problems is well known, little research toward solving them exists. Consequently, researchers know little about how the multicast infrastructure operates, how to solve existing problems, and what other problems exist. This article seeks to remedy this lack of knowledge by analyzing multicast routing instabilities at individual routers and comparing views between them. Our goal is to present a network-layer view of multicast routing stability and to characterize instabilities by evaluating temporal and spatial variations in the routing views. We also explore potential causes of these instabilities.