Abstract: The parallelizing of relational operations, especially join operation, has been a key research issue in parallel databases. Most of researchers agree that the Shared-Nothing architecture is most suitable to very large parallel databases.So, this paper presents three parallel join algorithms based on Snared-Nothing multiprocessor system and compares their performance. Finally, some open problems arc introduced.

摘要： 在并行数据库中,各种关系操作,特别是连接(Join)操作的并行化一直是一个研究热点。一般认为,无共享(Shared-Nothing)结构是最适于超大规模并行数据库的并行体系结构。因此,本文主要介绍2种基于无共享多处理机系统的并行连接算法,对它们的性能作了比较,并指出在该领域仍存在的一些主要问题。

关键词: 并行数据库, 并行体系结构, 无共享, 连接, 分类—归并, Gracc散列, Hybrid散列

CLC Number: 

• TP338.6