测控消息中间件发布/订阅匹配算法

Publish/Subscribe Matching Algorithm for Message-oriented Middleware

针对传统谓词计数法和谓词索引法在进行测控消息发布与订阅过程中效率较低的问题,结合目前测控计算机系统信息交换特点,提出一种基于谓词划分的发布/订阅匹配算法。预处理时将订阅按主题分类,同一主题下的所有订阅组织成0,1矩阵,矩阵每行对应一个订阅,每列对应一个谓词,不同行、列分别按所含谓词、所属订阅数量从大到小排列。匹配时若事件属性与某个谓词不匹配,则删除包含此谓词和其覆盖谓词的所有订阅,剩余矩阵行对应订阅即为匹配订阅集。算法匹配过程有效减少了谓词匹配量,时间复杂度达到线性。实验结果表明,当同一主题订阅量为40/s以下时,该算法匹配效率比传统谓词计数法和索引法有较大提高。

Aiming at the problem that the efficiency with traditional predicate counting method and predicate index algorithm is low when doing the publishing and subscribing for telemetry and command computer messages. Combined with the message exchange characteristics of current telemetry and command computer,a subscribe/publish matching algorithm is put forward based on predicate division. In this al-gorithm,all subscriptions are classified by topic in preprocessing,subscriptions under one topic are organized into matrix of （0,1）,each row corresponding to one subscription,each column corresponding to one predicate. Different rows and columns are rearranged by de-scending order according to the number of corresponding predicates and subscriptions. When an attribute does not matched with one predi-cate,all subscriptions containing the predicate and the ones covered by the predicate shall be deleted and the subscriptions corresponding to the reaming rows are the matched ones. Experimental results show that the matching efficiency of matrix matching algorithm is higher than traditional predicate matching algorithm and violent matching when the number of subscriptions under one topic is below 40 per sec-ond.