摘要： 针对分布式数据库系统在使用分配算法时需要预先指定事务执行站点的情况,通过分析分布式数据库查询代价模型,提出一种事务分类部署策略。利用层次聚类算法对查询事务进行分类并将同类事务部署至同一站点,在聚类过程中为查询事务构造查询矩阵和相似度矩阵,降低查询事务执行时间。实验结果表明,在相同的分片冗余条件下,当测试表数据量大于100万条且并发人数为100人时,与集中式数据库系统相比,基于该策略的分布式数据库系统具有更短的并发查询时间和更快的系统响应速率。

关键词: 分布式数据库, 层次聚类算法, 查询矩阵, 相似度矩阵, 事务聚类, 并发查询时间

Abstract: In view of the problem that the distributed database system needs to specify the execution site for transactions in advance,this paper proposes a transaction classification deployment strategy of distributed database by analyzing the distributed database query cost model.The similar transactions which are clustered by hierarchical clustering algorithm are deployed to the same site.Query matrix and similarity matrix are constructed for query transaction in the clustering process.Experimental results show that under the conditions that each slice has the same redundancy,test table data are more than 1 million and the number of concurrent people is 100,the distributed database system based on this strategy has shorter concurrent query time and faster system response rate than centralized database system.

Key words: distributed database, hierarchical clustering algorithm, query matrix, similarity matrix, transaction clustering, concurrent query time

中图分类号: 

• TP311