基于相似树的MPLS多播标记聚合算法

为了减少MPLS网络多播标记的占用空间,本文提出了一种基于相似树的求解算法,将与同一共用树相似的多播组进行标记聚合。首先运用模糊聚类的方法生成共用树,然后以共用树为基础进行多播树的构建和组成员的更新。动态多播组成员的变化大多只会造成LER节点类型的转换,不需要重新分配标记。实验结果表明,本算法可以大大缩小MPLS标记的占用,在 元素数量大于最大数量的1/2、相似度阈值 =0.75时,能够将大于50%的多播组进行标记聚合,而平均每组多用的节点数小于0.5。

A Label Aggregation Algorithm for MPLS Multicast Based on Similar Tree

In this paper, a similar tree-based algorithm is proposed to reduce the label space of multicast in multi-protocol label switching (MPLS) networks. The proposed algorithm aggregates the labels of multicast groups that are similar to the same common tree. In this algorithm, common trees are created via fuzzy clustering, based on which multicast trees are built and group members are updated. The change of dynamic multicast memberships can only give rise to the type conversion of label edge router (LER) nodes, without redistributing labels for the groups. Test results indicate that the proposed algorithm greatly reduces the quantity of MPLS labels being used, and that, when the number of λ element is greater than half of the maximum and when the similarity threshold equals 0.75, more than 50% of the multicast groups can be aggregated, with an average waste of less than 0.50 nodes in each group.