中文摘要：

为支持新兴网络应用，IP组播（multjcasc）和MPLS（multi-protocollabelswja。hing）技术分别从不同方向扩展了当前的IP路由和交换模式．MPLS和IP组播的结合是当前研究的一个热点，MPLS网络中的服务质量组播面临着标签资源匮乏、组播路由状态的可扩展性以及具体实现上的困难．针对这些问题，提出了基于汇聚方法的新型MPLS服务质量组播体系结构，提出在现有的路由控制平面上叠加一层面向IP组播服务的控制平面，取代组播路由协议并支持组播聚集，形成2层控制平面结构．定义了两平面之间的协作和交互方式，并通过扩展RSVP—TE（resourcereservalionprotocol—trafficengineering）P2MP（pointtomulti-point）协议，在新的体系结构中融合了服务质量控制能力．另外，还探讨了汇聚组播中基于距离约束选择汇聚路由器的算法，实现了基于Linux的MPLS组播路由器和IP组播服务控制系统，并组建了实验平台．实验和模拟结果表明，基于汇聚组播的双平面网络控制结构能够适应组播用户和网络拓扑的动态变化，能够有效节省MPLS标签资源，平衡网络中组播流量的分布．

英文摘要：

IP multicast and MPLS （multi-protocol label switching） are proposed to support emerging network applications effectively, extending current IP routing and switching mode in different ways. It is still in progress to combine IP multicast with MPLS properly. In this paper, a novel architecture, called Rendezvous MPLS Multicast, is proposed to support IP multicast in MPLS networks and to have the scalability of multicast routing, to achieve the label space reduction, and to solve some practical problems in implementation. A two-layered structure of control plane without multicast routing protocols is implemented by overlaying novel service control plane over the existing routing control plane to support multicast state aggregate. The interface between the two planes is formulated to support the interaction and cooperation. Moreover, the label distribution process for MPLS P2MP （point to multi-point） connection, with RSVP-TE （resource reservation protocol-traffic engineering） protocol, is extended to support aggregate of multiple label switching paths with traffic engineering and guarantee of quality of service. An algorithm of selecting Rendezvous Routers for Rendezvous MPLS Multicast has been presented, A test-bed with Linux-based implementations of MPLS multicast router and IP multicast service control system has also been constructed. The experimental results show its efficiency in terms of label space reduction and multicast traffic balancing, with the whole system adapting to the dynamic change of group members and network topology.