资源描述:
《基于Authorware的交互课件的设计与制作.doc》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、TheHadoopDistributedFilesystemWhenadatasetoutgrowsthestoragecapacityofasinglephysicalmachine,itbecomesnecessarytopartitionitacrossanumberofseparatemachines.Filesystemsthatmanagethestorageacrossanetworkofmachinesarecalleddistributedfilesystems.Sincetheyarenetwork-based,allthec
2、omplicationsofnetworkprogrammingkickin,thusmakingdistributedfilesystemsmorecomplexthanregulardiskfilesystems.Forexample,oneofthebiggestchallengesismakingthefilesystemtoleratenodefailurewithoutsufferingdataloss.HadoopcomeswithadistributedfilesystemcalledHDFS,whichstandsforHado
3、opDistributedFilesystem.(Youmaysometimesseereferencesto“DFS”—informallyorinolderdocumentationorconfiguration—whichisthesamething.)HDFSisHadoop’sflagshipfilesystemandisthefocusofthischapter,butHadoopactuallyhasageneralpurposefilesystemabstraction,sowe’llseealongthewayhowHadoop
4、integrateswithotherstoragesystems(suchasthelocalfilesystemandAmazonS3).TheDesignofHDFSHDFSisafilesystemdesignedforstoringverylargefileswithstreamingdataaccesspatterns,runningonclustersoncommodityhardware.Let’sexaminethisstatementinmoredetail:Verylargefiles“Verylarge”inthiscon
5、textmeansfilesthatarehundredsofmegabytes,gigabytes,orterabytesinsize.ThereareHadoopclustersrunningtodaythatstorepetabytesofdata.*StreamingdataaccessHDFSisbuiltaroundtheideathatthemostefficientdataprocessingpatternisawrite-once,read-many-timespattern.Adatasetistypicallygenerat
6、edorcopiedfromsource,thenvariousanalysesareperformedonthatdatasetovertime.Eachanalysiswillinvolvealargeproportion,ifnotall,ofthedataset,sothetimetoreadthewholedatasetismoreimportantthanthelatencyinreadingthefirstrecord.CommodityhardwareHadoopdoesn’trequireexpensive,highlyreli
7、ablehardwaretorunon.It’sdesignedtorunonclustersofcommodityhardware(commonlyavailablehardwareavailablefrommultiplevendors†)forwhichthechanceofnodefailureacrosstheclusterishigh,atleastforlargeclusters.HDFSisdesignedtocarryonworkingwithoutanoticeableinterruptiontotheuserinthefac
8、eofsuchfailure.ItisalsoworthexaminingtheapplicationsforwhichusingHDF