Scalable Wide Area Multicast Routing

[406007 . Estrin]该提案首先简要回顾了组播路由的发展，并确定了在全球规模的互联网中引入组播服务所引起的扩展问题。然后，我们提出了评估缩放问题的不同解决方案的标准。根据这些评估标准，我们概述了我们提出的解决方案，称为PIM。PIM可以显著提高组播路由的可扩展性，并且能够提供比其他可扩展组播方案（如基于核心树）更好的服务质量。我们给出了初步的仿真结果，比较了不同分布树的性能。然而，在非常大的异构网络中充分表征组播路由方案的性能并非易事；在我们的目标环境（即全球互联网）规模的运营网络中，“通过经验学习”是不可行的。描述组播协议不仅具有挑战性，因为目标互联网的规模，而且还因为协议本身的复杂性和灵活性。此外，单个多播协议（如PIM）可以以各种直接影响性能的方式使用（或配置）。因此，为了描述我们提出的多播方法，并相对于替代方案评估它，我们提出了广泛的模拟研究。提出的研究由三部分组成：1)广泛的模拟研究，以表征所提出的机制所呈现的缩放特性和权衡；2)研究替代聚合机制，以进一步扩展可扩展性（使用我们构建的模拟工具进行评估）；3)在实验室和测试平台网络中测试所提出方案的原型实现。＊＊＊

9406007 Estrin The proposal begins with a brief review of the evolution of multicast routing and identification of the scaling issues raised by the introduction of multicast services in a global-scale internet. We then present criteria for evaluating different solutions to the scaling issues. Given these evaluation criteria we outline our proposed solution, called PIM. PIM could significantly improve the scaling properties of multicast routing and is capable of providing better service quality than alternative scalable multicast schemes such as Core Based Trees. We present preliminary simulation results comparing the performance of different distribution trees. However, fully characterizing the performance of multicast routing schemes in very large, heterogeneous, internetworks is not trivial; and ``learning through experience'' in operational networks that are of the size of our target environment (i.e., the global internet), is not feasible . Characterizing multicast protocols is not only challenging because of the scale of the target internet, but also because of the complexity and flexibility built into the protocols themselves. Moreover, a single multicast protocol such as PIM, may be used (or configured) in a variety of ways that directly impact performance. Therefore, to characterize our proposed multicast approach, and to evaluate it relative to alternatives, we propose extensive simulation studies. The proposed research is composed of three parts: 1) an extensive simulation study to characterize the scaling properties and tradeoffs presented by the proposed mechanisms, 2) an investigation of alternative aggregation mechanisms to further extend scalability (using our constructed simulation tools for their evaluation), and, 3) a prototype implementation of the proposed scheme for testing in laboratory and testbed networks. ***