摘要： 提出了依照并行搜索树的思想，将订阅的属性集划分为2个属性子集，在2个子集上分别构建一个并行搜索树。匹配事件时，在两个搜索树上分别得到该事件匹配的订阅集，两个集合的交集即为最终结果。分析及实验结果表明：划分并行搜索树可以减少搜索树的节点数，降低对存储空间的需求，并潜在地缩短事件匹配时间，尤其适合于节点分支数多而匹配分支少时的情况。

关键词: 发布/订阅系统, 事件匹配, 并行搜索树

Abstract: This paper presents the parallel search tree based on attribute set division (PST-ASD). It divides the attribute set into two attribute subsets, and constructs a parallel search tree (PST) respectively on these two attribute subsets. When matching an event, it gets the matching subscription sets from the two search trees respectively, and then gets the finally result by intersecting these two sets. Analysis and experiments show that PST-ASD takes on much less store space and in certain conditions takes on less matching time, especially in the condition that the average number of branches following each tree node is large while the number of matching branches is small.

Key words: Publish/subscribe systems, Event matching, Parallel search tree