面向分布式的SPARQL查询优化算法  

Distributed Optimized Query Algorithm Based on SPARQL

作　　者：汪璟玢[1] 方知立 张燕琴[1] 

机构地区：[1]福州大学数学与计算机科学学院,福州350108

出　　处：《计算机科学》2014年第7期227-231,共5页Computer Science

基　　金：福州大学科技发展基金资助项目(2013-XQ-32);空间数据挖掘与信息共享教育部重点实验室开放研究基金项目(201006);2011年福建省科技拥军基金项目(JG2011005);福建省自然科学基金项目(2012J01168)资助

摘　　要：采用分布式来实现SPARQL(Simple Protocol and RDF Query Language)查询是解决海量RDF(Resource Description Framework)查询的一种新思路。目前实现的基于Hadoop的RDF查询都要启用多个MapReduce来完成任务,浪费时间。为了克服此缺点,提出MRQJ(using MapReduce to query and join)算法,用以实现SPARQL的分布式查询。该算法分为连接计划生成与SPARQL查询执行两个部分:连接计划生成采用贪心策略,生成最优的连接方案;在SPARQL查询执行中只需结合一次MapReduce计算即可得到查询结果。在LUBM数据集上进行的测试实验表明:在查询语句较为复杂的情况下,MRQJ方法的查询效率具有明显的优势。It＇s a new way of solving the large amount of RDF（Resource Description Framework） query problem to use distributed technique to realize the SPARQL（Simple Protocol and RDF Query Language） Query. At present, most of the RDF queries based on Hadoop have to use multiple MapReduce jobs to complete the task, resulting in waste of time. In order to overcome this drawback, this paper proposed MRQJ（using MapReduce to the query and the join） algo- rithm to perform distributed SPARQL query. The algorithm can be divided into join plan generation and SPARQL que- ry execution two parts ：join plan generation uses greedy strategy to generate the most optimal join scheme, and only one MapReduce job should be done to get the query results in SPARQL query execution. The experiment on the LUBM test data set was made. The experimental results show that MRQJ method query efficiency is higher when the case of a que- ry is more complicated.

关 键 词：RDF Hadoop SPARQL查询 MAPREDUCE 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]