一种改进的多维度并行匹配发布与订阅算法

An Improved Multi-Dimensional Parallel Matching Publish and Subscribe Algorithm

基于内容的发布订阅是一种比基于主题的发布订阅更加灵活的解耦通信方式,传统的基于内容的匹配算法中大多基于单维度以流水线的方式匹配,如REIN[1]。随着数据规模的增长,这些算法由于串行搜索的限制逐渐不能满足高性能要求。提出一种基于内容的面向大规模发布订阅场景的匹配算法MP-REIN,改进其单维度匹配为多维度并行匹配,考虑使用多线程提高匹配算法的处理速度,使用数量可控的线程并行处理订阅空间内所有属性,大幅度提高了事件匹配的速度。实验结果显示MP-REIN算法在匹配时间和插入时间上较传统的算法有一定的提升。

Content-based publish and subscribe is a more flexible decoupling communication method than topic-based publish and subscribe.Tradi⁃tional content-based matching algorithms are mostly based on single-dimensional matching in a pipelined manner,such as REIN[1].With the increase of data size,these algorithms can not meet the high performance requirements due to the limitation of serial search.This paper proposes a content-based matching algorithm for large-scale publish and subscribe scenarios,MP-REIN,which improves its single-di⁃mensional matching to multi-dimensional parallel matching.It considers using multi-threading to improve the processing speed of the matching algorithm and uses a controlled number of threads in parallel Processing all attributes in the subscription space,greatly improving the speed of event matching.The experimental results show that the MP-REIN algorithm has a certain improvement over the traditional al⁃gorithm in terms of matching time and insertion time.