基于Hadoop的QR树索引方法  

Hadoop-based QR-tree index

作　　者：冯钧[1] 任锋[1] 唐志贤[1] 

机构地区：[1]河海大学计算机与信息学院,江苏南京211100

出　　处：《计算机工程与设计》2013年第12期4231-4236,共6页Computer Engineering and Design

基　　金：江苏省科技支撑计划(工业)基金项目(BE2012179);江苏省普通高校研究生科研创新计划基金项目(CXZZ12_0229)

摘　　要：针对处理海量空间数据时,集中式环境在机器性能上无法满足超大计算量的性能要求问题,提出基于分布式平台Hadoop来替换集中式环境,突破机器性能瓶颈,并且利用QR-Tree为海量空间数据建立索引,同时参考MapReduce"分而治之"的思想,设计了基于MapReduce并行框架处理索引创建和查询的算法。通过在分布式环境下改造QR-Tree算法为HQR-Tree(Hadoop QR Tree)化整为零,将计算量分散到Hadoop各个计算节点中并行计算,以提高计算效率,减少响应时间。实验结果表明,HQR-Tree处理海量空间数据具有较高的效率。For dealing with massive spatial data, centralized environment on the performance of the machine can not meet the large computational requirements, a method is proposed to replace the centralized environment based on Hadoop to break through the bottleneck of machine performance, and the QR-Tree is used for massive spatial data indexing, MapReduce thinking of ＂di- vide and rule＂ is referenced, creating and query-based MapReduee parallel framework is designed to handle Index algorithm. Transforming QR-Tree algorithm in a distributed environment to the HQR-Tree （Hadoop QR-Tree）, and in order to improve the computational efficiency and reduce response time, the calculation amount is distributed to the slaves. Experiments show the HQR-Tree can handle massive spatial data with high efficiency.

关 键 词：QR树 空间数据索引 分布式数据索引 HADOOP MAPREDUCE 

分 类 号：TP311.131[自动化与计算机技术—计算机软件与理论]