RR_tree:多维索引的关系模式实现新方法

RR-tree:A New Relational Approach of Multi-dimension Index

为了有效地管理多媒体信息、地理信息及空间数据,提出了多种针对多维数据的索引方法。其中一些索引方法已经在现有的商用数据库系统(DBMS)得以实现,然而学术研究及实际应用中需要更多种的多维乃至高维数据索引方法的支持。有研究提出在关系数据库上利用存储结构、存储过程、触发器来模拟并实现X_tree的多维数据索引功能。在此基础上加以改进,重新设计了模式结构,增加了关键的索引,引入了聚簇存储,以关系模式实现多维索引的创建、插入、查询等操作;并且与现有的商用数据库系统的多维索引Oracle Spatial进行了插入、查询的性能比较。实验结果充分证明这种以关系模式实现多维索引方法的可行性与可用性。

A number of studies have reported various methods for indexing multi-dimensional data, which is important to manage multimedia information and GIS spatial data. The existing commercial DBMSs （database management systems） have already implemented some of those methods. However, contrasted with the academic research, practical applications need more kinds of index methods for the multi-dimensional even high-dimensional data. To actualize the multi-dimensional model for index creation, insertion, query and other operations in DBMS more effectively and efficiently, this paper makes use of data storage structure, stored procedures and triggers of relational database to simulate R_tree, a popular index method of multi-dimensional data. The method is named as RR_tree, a new relational approach of multi-dimensional index. The performance experiments of insertion and query show that, compared with the Oracle Spatial index, the existing commercial multi-dimensional database system, the RRtree is feasible and usable in relational realization of the multi-dimensional model.