VPLS中基于共享聚合树的组播问题研究

Study on the Delivery of Multicast Traffic in VPLS Using Aggregated Multicast Trees

VPLS是近年来网络研究的热点.但在设计VPLS网络的时候,面临着一个严峻的挑战组播中需要维护的状态数和网络带宽浪费之间的矛盾问题.本文提出了一个基于共享聚合树的解决方案,既减少了需要维护的组播树数目,又较好地控制了带宽的使用.其基本方法是,采用Prim贪婪算法建立洪泛树,确保组播流量能够传送到所有的PE结点.在洪泛树的基础上通过基于消息的剪枝机制建立候选的组播树,并通过组-树之间的映射算法实现组播树的共享和聚合.候选树的产生是通过预设的带宽阀值来触发的.仿真实验表明,该方法具有较好的性能,能够降低VPLS网络中由于维护组播转发状态而带来的开销.

VPLS has always been the focus in today＇s network research. But implementing multicast in VPLS is confronted with two major problems, one is the multicast forwarding states that routers must maintain and manage, the other is the bandwidth waste when using aggregate shared trees to deliver multicast traffic, especially when multimedia traffic delivered along the tree. This paper proposes a trade-off approach that can be used to reduce the multicast forwarding states while at the same time keep the bandwidth utilized efficiently. It begins with the set up of aggregate default tree, using the Greedy-prim algorithm to build the minimal spanning tree with the useless nodes removed. With the aggregate default tree, candidate shared multicast trees are derived, which is selected if its bandwidth waste is below a certain pre-determined threshold. The group-tree selection algorithm is also described and fully illustrated. Simulation result shows that the proposed approach in this article can achieve better scalability performance but introduce little overhead on existing PE or P touters in VPLS networks.