A hierarchical architecture to support large-scale application-layer multicast.

详细信息   

• 作者：Jiang ; Yingyin.
• 学历：Doctor
• 年：2005
• 导师：Wu, Min-You
• 毕业院校：The University of New Mexico
• 专业：Engineering, Electronics and Electrical.;Computer Science.
• ISBN：9780542263996
• CBH：3184435
• Country：Mexico
• 语种：English
• FileSize：8503448
• Pages：184

文摘

The ubiquitous network has connected the world altogether wherever you are sitting. However, people is no longer satisfied with only point-to-point communication. For more than a decade, multi-point communication has become a hot topic in the research community. Besides the efforts on IP multicast, the application-layer or overlay multicast recently has received more attention. In this dissertation, I propose a multicast architecture where multi-point delivery is viewed as an application-level service. The architecture relies on a collection of strategically placed proxies that collaboratively provides the multicast service. Clients locate a preferred parent proxy and join the multicast session via that preferred proxy. Proxies organize themselves into an overlay network through unicast connections and build data distribution trees on top of this overlay structure.;Most of the recently proposed overlay multicast solutions do not address the inter-domain administrative issue. For wide-area multicast, it is unavoidable to have communication among several administrative domains. At this time, multicast routing cannot assume the full knowledge of global network topology and ignore the routing policy constrained by the contractual commercial agreement between administrative domains. This problem can be solved by a hierarchical organization of the routing infrastructure. In this work, I propose a new approach for overlay multicast that performs hierarchical routing on two levels---one level for within each of the administrative domains and another level for among the domains. I compare a single, centralized routing to the zone-approach routing for overlay multicast; the results have shown that the distributed design can achieve very close performance to the centralized algorithm. At the same time, the hierarchy enables the composition of different administrative networks, each with its own independent multicast protocol and administrative policy.;The second part of the architecture is a Mechanism Design approach for overlay multicast routing. Traditional multicast routing protocol assumes the entities participating in the routing follow the prescribed algorithm without any deviation. This common belief cannot hold if we consider each administrative region, which belongs to different private organizations, as rational, selfish and making its own decision. Although multicast saves resource consumption for the whole system, it does not necessarily benefit every intermediate agent that makes a replication. As an agent responds to incentive, only when a zone can benefit from replicating packets will it join the multicast session and contribute. (Abstract shortened by UMI.)