A Scalable and Robust Multicast Architecture for the Internet

点击翻译按钮获取中文摘要

A critical challenge for development of the Internet is how to support efficient and effective multicast communication in response to the intense pace of growth of the Internet and the concomitant demand for communication services from its users. Current multicast protocols, while well-suited to local or smaller scale communication, have suffered from serious shortcomings with respect to scalability, stability, and inter-domain coordination. Our approach to the multicast routing problem involves the introduction of new graph theoretic structures as underlying abstract communication topologies for multicast and the integration of these new structures with current technologies. There are two key elements to our new multicast architecture: a) 3-layer hierarchical multicast protocol and b) a virtual topologies model. Together these two elements provide a flexible architecture for multicast that has the potential to overcome the shortcomings described above. In this proposal, we focus on the virtual topologies element of the architecture. Certain graph structures have great potential to serve as underlying topologies for multicast communication, due to properties of connectivity, symmetry, limited vertex degree, and small diameter. Under the proposed model, an abstract communication topology, such as the k-tree or the k-ary n-cube, is used as basis for the execution and routing of the multicast protocols. The mapping of the virtual topology to physical routes in the Internet occurs dynamically, as is done under current, shared-tree approaches to multicasting. More robust topologies can be expected to provide greater flexibility and fault tolerance than simple tree structures. Scalability comes from the fact that these virtual topologies are limited to intradomain application. Our proposed 3-level, hierarchical multicast protocol follows the Internet domain structure, interconnecting domain-specific virtual topologies into an overall multicast structure. We plan to develop new pr otocols for multicast under the architectural model described above and to test their feasibility and performance through simulation. Our interdisciplinary team of researchers in graph theory, parallel computing, and networking brings a new synergy to the multicast problem. Our theoretic work will be guided by an understanding of the pragmatic issues involved in today's Internet protocols. We hope to define and develop fundamental new multicast structures and algorithms that can ultimately serve as the foundations for a realistic, effective multicast architecture for the Internet.