资源描述:
《基于分组序号的聚集算法pdf》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、1000-9825/2003/14(02)0222©2003JournalofSoftware软件学报Vol.14,No.2∗基于分组序号的聚集算法1+12冯建华,蒋旭东,孟宪虎1(清华大学计算机科学与技术系,北京100084)2(运城高等专科学校计算机系,山西运城044000)AnAggregationAlgorithmBasedonGroupNumbers1+12FENGJian-Hua,JIANGXu-Dong,MENGXian-Hu1(DepartmentofComputerScienceandTec
2、hnology,TsinghuaUniversity,Beijing100084,China)2(DepartmentofComputer,YunchengAdvancedTrainingCollege,Yuncheng044000,China)+Correspondingauthor:Phn:86-10-62789150,E-mail:fengjh@tsinghua.edu.cnhttp://www.tsinghua.edu.cnReceived2001-10-22;Accepted2002-04-22Fe
3、ngJH,JiangXD,MengXH.Anaggregationalgorithmbasedongroupnumbers.JournalofSoftware,2003,14(2):222~229.Abstract:OLAP(onlineanalyticalprocessing)queriesarecomplex.WhenimplementedinSQL(structuredquerylanguage),theyusuallyinvolvemulti-tablejoinandaggregateoperatio
4、ns.Asaresult,howtoimprovetheperformanceofthemulti-tablejoinandaggregateoperationsbecomesakeyissueforROLAP(relationalOLAP)queryevaluation.Tosolvethisproblem,anaggregationalgorithmbasedongroupnumbersnamedMuGA(groupnumberbasedaggregationwithmulti-tablejoin)isp
5、roposedinthispaper.Bytakingthecharacteristicsofstarschemaintoconsideration,thealgorithmcombinestheaggregationoperationwiththenovelmulti-tablejoinalgorithm,Mjoin(multi-tablejoin),andreplacesthesortingandhashingmethodbycomputedgroupnumbersinaggregationcomputi
6、ng.Asaresult,thealgorithmcannotonlyreducetheCPUtime,butalsoreducethediskI/OsforOLAPqueries.Asillustratedbytheexperiments,theperformanceofthealgorithmMuGAissuperiortooriginalaggregationmethodsandthenewsortingbasedmethodforaggregation.Keywords:datawarehouse;O
7、LAP(onlineanalyticalprocessing);multi-tablejoin;aggregationquery摘要:联机分析处理OLAP(onlineanalyticalprocessing)查询作为一种复杂查询,当使用SQL(structuredquerylanguage)语句来表述时,通常都包含多表连接和分组聚集操作,因此提高多表连接和分组聚集计算的性能就成为ROLAP(relationalOLAP)查询处理的关键问题.提出一种基于分组序号的聚集算法MuGA(groupnumberbas
8、edaggregationwithmulti-tablejoin),该方法充分考虑数据仓库星型模式的特点,将聚集操作和新的多表连接算法MJoin(multi-tablejoin)相结合,使用分组序号进行分组聚集计算,代替通常的排序或者哈希计算,从而有效地减少CPU运算以及磁盘存取的开销.算法的实验数据表明,提出的MuGA算法与传统的关系数据库聚集查询∗SupportedbytheNationalGr