欢迎来到天天文库
浏览记录
ID:9853319
大小:127.50 KB
页数:25页
时间:2018-05-12
《全文搜索引擎的设计与实现-外文翻译》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、江汉大学毕业论文(设计)外文翻译原文来源TheHadoopDistributedFileSystem:ArchitectureandDesign中文译文Hadoop分布式文件系统:架构和设计姓名XXXX学号72013年4月8日英文原文TheHadoopDistributedFileSystem:ArchitectureandDesignSource:http://hadoop.apache.org/docs/r0.18.3/hdfs_design.htmlIntroductionTheHadoopDistributedFileSystem(HDFS)isadistributedfile
2、systemdesignedtorunoncommodityhardware.Ithasmanysimilaritieswithexistingdistributedfilesystems.However,thedifferencesfromotherdistributedfilesystemsaresignificant.HDFSishighlyfault-tolerantandisdesignedtobedeployedonlow-costhardware.HDFSprovideshighthroughputaccesstoapplicationdataandissuitablef
3、orapplicationsthathavelargedatasets.HDFSrelaxesafewPOSIXrequirementstoenablestreamingaccesstofilesystemdata.HDFSwasoriginallybuiltasinfrastructurefortheApacheNutchwebsearchengineproject.HDFSispartoftheApacheHadoopCoreproject.TheprojectURLishttp://hadoop.apache.org/core/.AssumptionsandGoalsHardwa
4、reFailureHardwarefailureisthenormratherthantheexception.AnHDFSinstancemayconsistofhundredsorthousandsofservermachines,eachstoringpartofthefilesystem’sdata.Thefactthatthereareahugenumberofcomponentsandthateachcomponenthasanon-trivialprobabilityoffailuremeansthatsomecomponentofHDFSisalwaysnon-func
5、tional.Therefore,detectionoffaultsandquick,automaticrecoveryfromthemisacorearchitecturalgoalofHDFS.StreamingDataAccessApplicationsthatrunonHDFSneedstreamingaccesstotheirdatasets.Theyarenotgeneralpurposeapplicationsthattypicallyrunongeneralpurposefilesystems.HDFSisdesignedmoreforbatchprocessingra
6、therthaninteractiveusebyusers.Theemphasisisonhighthroughputofdataaccessratherthanlowlatencyofdataaccess.POSIXimposesmanyhardrequirementsthatarenotneededforapplicationsthataretargetedforHDFS.POSIXsemanticsinafewkeyareashasbeentradedtoincreasedatathroughputrates.LargeDataSetsApplicationsthatrunonH
7、DFShavelargedatasets.AtypicalfileinHDFSisgigabytestoterabytesinsize.Thus,HDFSistunedtosupportlargefiles.Itshouldprovidehighaggregatedatabandwidthandscaletohundredsofnodesinasinglecluster.Itshouldsupporttensofmillionsoffilesi
此文档下载收益归作者所有