数据仓库物化视图在线算法研究

《数据仓库物化视图在线算法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、东南大学博士学位论文数据仓库物化视图在线算法研究姓名：张柏礼申请学位级别：博士专业：计算机应用技术指导教师：孙志挥20050701AbstractAdatewarehouseisadatarepository,whichcollectsandmaintainsalargemountofdatafrommullipledism'buted，autonomousandpossiblyheterogeneousdatasoIM．Oftenthedataisstoredintheformofmaterializedview3forthepurposeefficientlyimpl

2、ementingdccision-suppo仳orOLAPqueries．Firstly,theselectionofviewsformaterializationisoneofthemostissuesinthedesigndatewarehouse,it’sgoalistoselect∞appropriatesetofviewsthatminimizestotalqueryresponsetimeand／orthecostofmaintainingtheselectedviews,givonalimitedamountofrasonsce，e．g．，storagesp

3、ace，materializationtime，ortotalviewmaIntanancetime．Secondly,maintaininglotsofmaterializedviewsischallenging．especiallyffthedatasources2reautonomousandviewsofthedataatthedatawarehousespanmultiplesoume“q．Theviewshavetobemaintainedintimetoreflecttheupdatesdoneagainstthebaserelationsstoredatt

4、hevariousdatesources．n圮e伍cientmaintenanceofmaterializedviewshasbecomeanotherimportantresearchiss∞．Theusualmethodsis廿扭tthematerializedviewsaraselectedandmaintainedoffline。forexample,thestaticselectionalgorithmofviewisemployedtogenerateorregeneratethematerializedviewset，andbatchupdatepolicy

5、isusedtomaintainthematnight，However,wheremtheamountofdataenteringawerchousc,thequeryloads，andtheneedtoobtain叩-to-dateresponsesareauincreasing,the‘'night"’timeavailableformakingthewarehouseup-to-dateisshrinking．Theseu'andsnecessitateefficientonlinealgorithmsformaterializedviews．Inthispaper

6、,itisemphasizedtomakedeeperresearchontheselectionalgorithmsandmaintenancealgorithmsofmaterializedviewsduringthedatawarehouseon-line．Themaincontributionsofthisdissertationareasfollows：(1)Tosolvetheshortageofthestaticselectionalgorithmofmaterializedview,thepaperpresentsPMVS．preprocessorofma

7、terializedviewselectionapproach．PMVSmakesu∞ofthequerysparsenessanddatasparsenassinmulti-dimensionaldatasets，andreducesthetimecomplex(searchspacc)ofthestaticselectionalgorithm，Sothecostofstaticalgorithmsonspaceandtimecanbecutdowntofitforonlinedemand．(2)Preseataimprov