Service overlay networks for multicast applications
Mário Gerla, Li Lao · 2006
The growing popularity of the Internet as a pervasive communication media has led to an increasing usage for group-oriented applications such as IPTV, online gaming and video conferences. To support such applications, IP multicast has been proposed by Deering in 1988. However, IP multicast faces many technical and marketing problems, such as deployment difficulties, the state scalability issue and the lack of good pricing models. Recently, application layer multicast and overlay multicast have emerged as alternative solutions. Although they address some of the above issues, there are fundamental limitations in both solutions, especially with respect to scalability issues when a multicast group is very large or when there are many co-existing groups. In addition, multicast research has been traditionally plagued by a lack of systematic performance evaluation methodology. In this dissertation, our major contributions are: (1) an efficient hierarchical multicast service overlay architecture; (2) an analytic multicast performance evaluation framework; (3) an extensive simulation and experimental study to validate the results. We develop a comprehensive framework to systematically evaluate multicast performance. The framework takes into account realistic network topology and group membership models, and provides a rich set of configurable components for various multicast applications. Using this framework, we conduct a comparative study of the three multicast architectures using both qualitative and quantitative methods. Our study suggests that application layer multicast is suitable for small membership groups, but it is not efficient for large groups. On the other hand, we discover that by aggregating many groups in a multicast overlay (using proxies), we can achieve good performance. Inspired by this study, we propose a Two-tier Overlay Multicast Architecture (TOMA) to provide scalable and efficient multicast support. In TOMA, Multicast Service Overlay Network (MSON) is advocated as the backbone service domain, while end users in access domains form local clusters. We design a protocol OLAMP to effectively handle the state scalability problem. Extensive simulation studies demonstrate that TOMA can provide efficient resource utilization with less control overhead. It also significantly reduces multicast state and management overhead in the presence of multiple groups. In addition, we develop several provisioning algorithms to help MSON providers plan backbone service overlay, and suggest a cost-based pricing model to charge multicast groups. To verify the above results, we have implemented a prototype of TOMA in PlanetLab, an Internet-wide testbed. Experiment results indicate that our solution is ready for real deployment.