摘要： 在基于内容的发布/订阅系统中，每个订阅通常含有某几个属性，并且一些属性经常在订阅中同时出现。根据该情况，在一个多维结构和多个一维结构的基础上，提出一种基于属性分组的匹配结构。属性分组把所有事件模型中的属性分为多个组，每个组包含经常在订阅中同时出现的属性，基于此设计索引结构并构建覆盖网。实验结果表明，该匹配结构可在系统负载和带宽开销上取得较好的平衡。

关键词: 发布/订阅系统, 匹配结构, 属性分组, 系统负载, 带宽开销

Abstract: In the publish/subscribe system based on content, each subscription usually contains certain attributes, and some properties of the subscription often appears at the same time. According to the situation, on the basis of a single multi-dimensional structure and multiple single-dimensional structures, this paper proposes a matching structure based on attribute grouping. Attribute grouping divides the attributes in the event scheme into several groups, each contains attributes that appear together frequently in the subscriptions. It designs indexing structures and construct overlays based on attribute grouping. Experimental results show that matching structure can achieve better balance between the bandwidth cost and the storage load.

Key words: publish/subscribe system, matching structure, attribute grouping, system load, bandwidth cost

中图分类号: 

• TP393