欢迎来到天天文库
浏览记录
ID:351236
大小:76.50 KB
页数:10页
时间:2017-07-26
《Delta-encoding与相似检测应用说明【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、本科毕业论文(设计)外文翻译外文原文Application-specificDelta-encodingviaResemblanceDetectionFredDouglisIBMT.J.WatsonResearchCenterHawthorne,NY10532douglis@acm.OrgArunIyengarIBMT. J.WatsonResearchCenterHawthorne,NY10532aruni@us.ibm.comAbstract Manyobjects,suchasfiles,electronicmessages,andwebpages,conta
2、inoverlappingcontent.Numerouspastresearchprojectshaveobservedthatonecancompressoneobjectrelativetoanotheronebycomputingthedifferencesbetweenthetwo,butthesedelta-encodingsystemshavealmostinvariablyrequiredknowledgeofaspecificrelationshipbetweenthem.Mostcommonly,twoversionsusethesamename
3、atdifferentpointsintime.Weconsidercasesinwhichthisrelationshipisdetermineddynamically,whenasufficientresemblanceexistsbetweentwoobjectsinarelativelylargecollection.Welookatspecificexamplesofthistechnique,namelywebpages,email,andfilesinafilesystem,andevaluatethepotentialdatareductionand
4、thefactorsthatinfluencethisreduction.Wefindthatdelta-encodingusingthisresemblancedetectiontechniquecanimprovesimplecompressionbyuptoafactoroftwo,dependingonworkload,andasmallfractionofobjectscanpotentiallyaccountforalargeportionofthesesavings.Keywords delta-encoding;ResemblanceDetectio
5、n1.IntroductionDelta-encodingistheactofcompressingadataobject,suchasafileorwebpage,relativetoanotherobject.Usuallythereisatemporalrelationshipbetweenthetwoobjects:thelatterobjectexists,andwhenitissubsequentlymodified,thechangescanberepresentedinasmallfractionofthesizeoftheentireobject.
6、Thereisoftenalsoanamingrelationshipbetweentheobjects,sinceamodifiedfilecanhavethesamenameastheoriginalcopy.Inthesecases,identifyingthebaseversionagainsttheonewhichtocomputeadeltaisstraightforward.Delta-encodingisparticularlyattractiveforthesituationswhereinformationisspreadingthroughan
7、etworkwithlimitedbandwidth.Forexample,websitesoftenreplicatethingsforhigherperformanceandavailability.Thebandwidthbetweenthereplicasislimited.Anotherexamplewouldbereplicatingmailsystems.Thesystemoftenallowsclientstoreplicatecopiesofmailmessageslocally.Clientscanconnectthetelephonenet
此文档下载收益归作者所有