一种面向海量分布式数据库的游标构造方法

摘要

游标是传统关系数据库所提供的一种能从包括多条数据记录的结果集中,每次提取一条记录的机制.这种机制能够缓存SQL查询结果集,提高重复查询的效率,并能以指针的方式,逐一访问缓存结果集的记录,非常灵活.但是,在现有的海量分布式数据库中,由于数据的分布式存储和数据量的庞大,游标的功能很难支持.文章提出了一种面向海量分布式数据库的游标构造方法,可以实现分布式数据库的游标功能,在研究了海量分布式数据库Oceanbase的基础上,构造了游标关键字共通的语法树、逻辑计划和物理计划;设计了游标整体的执行流程,并实现了游标功能.实验结果表明,在大数据量的查询过程中,文中所提出的游标构造方法所设计游标的查询速度,优于商业数据库DB2的游标查询速度.%Cursor is a mechanism provided by traditional relational databases to pick up one record each time from a result set which includes multiple records. This kind of mechanism can cache the SQL query result set, which speeds up the query efficiency, and can access each record in the cached result by the way of pointer flexibly. But in the existing massive distributed databases, cursor is very hard to support due to the distributed storage and large volume of data. This paper proposed a kind of cursor creating method for massive distributed database, which can realize the function of cursor. On the basis of studying the massive distributed database?Oceanbase, the method builds up the common syntax tree, logic plan and physical plan for the keyword of cursor, designs the execution procedure for cursor and implements the function of it. Experiments show that the proposed method is better than the commercial database?DB2 for cursor queries in big data.