资源描述：
《这个算是google云数据库的介绍吧》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。
1、BigTableASystemforDistributedStructuredStorageJeffDeanJointworkwith:OthertwelveguysCopyedfromBigTablevideotalkyfttyMotivationLotsof(semi-)structureddataatGoogleURLs:Contents,crawlmetadata,links,anchors,pagerank,…Per-userdata:Userpreferencesettings,recentqueri
2、es/searchresults,…Geographiclocations:Physicalentities(shops,restaurants,etc.),roads,satelliteimagedata,userannotations,…ScaleislargeBillionsofURLs,manyversions/page(~20K/version)Hundredsofmillionsofusers,thousandsofq/sec100TB+ofsatelliteimagedataWhynotjustus
3、ecommercialDB?ScaleistoolargeformostcommercialdatabasesEvenifitweren't,costwouldbeveryhighBuildinginternallymeanssystemcanbeappliedacrossmanyprojectsforlowincrementalcostLow-levelstorageoptimizationshelpperformancesignificantlyMuchhardertodowhenrunningontopof
4、adatabaselayerAlsofunandchallengingtobuildlarge-scalesystems:)GoalsWantasynchronousprocessestobecontinuouslyupdatingdifferentpiecesofdataWantaccesstomostcurrentdataatanytimeNeedtosupport:Veryhighread/writerates(millionsofopspersecond)Efficientscansoverallorin
5、terestingsubsetsofdataEfficientjoinsoflargeone-to-oneandone-to-manydatasetsOftenwanttoexaminedatachangesovertimeE.g.ContentsofawebpageovermultiplecrawlsBigTableDistributedmulti-levelmapWithaninterestingdatamodelFault-tolerant,persistentScalableThousandsofserv
6、ersTerabytesofin-memorydataPetabyteofdisk-baseddataMillionsofreads/writespersecond,efficientscansSelf-managingServerscanbeadded/removeddynamicallyServersadjusttoloadimbalanceStatusDesign/initialimplementationstartedbeginningof2004Currently~100BigTablecellsPro
7、ductionuseoractivedevelopmentformanyprojects:GooglePrintMySearchHistoryOrkutCrawling/indexingpipelineGoogleMaps/GoogleEarthBlogger…Largestbigtablecellmanages~200TBofdataspreadoverseveralthousandmachines(largercellsplanned)Background:BuildingBlocksBuildingbloc
8、ks:GoogleFileSystem(GFS):RawstorageScheduler:schedulesjobsontomachinesLockservice:distributedlockmanagerAlsocanreliablyholdtinyfiles(100sofbytes)w/highavailabilityMapReduce:simplifiedlarg
