文献翻译.docx
《文献翻译.docx》由会员分享,可在线阅读,更多相关《文献翻译.docx(5页珍藏版)》请在冰豆网上搜索。
![文献翻译.docx](https://file1.bdocx.com/fileroot1/2022-11/26/428f1d4c-8b78-41c4-954a-64a4e244c412/428f1d4c-8b78-41c4-954a-64a4e244c4121.gif)
文献翻译
河南科技学院
2016届本科毕业论文(设计)
中英文翻译
学生姓名:
XXXXXXX
所在学院:
信息工程学院
所学专业:
计算机科学与技术
导师姓名:
XXXXXX
完成时间:
2016年5月17日
Faultlocalizationbasedoncombinesactiveandpassivemeasurementsincomputernetworksbyantcolonyoptimization
Abstract:
Ascomputernetworkscontinuetogrowinsizeandcomplexity,effectivenetworkmanagementisexpectedtobecomeevenmorecruciallyimportantandmorechallenging.Computernetworkapplica-tionscanbeplaguedbyavarietyofsoftwareorhardwarefaults.Thesefaultscanbecriticalandcostlyinthedebugginganddeploymentofnetworks.Ingeneral,faultmanagementincomputernetworkscomprisesfoursteps:
faultdetection,faultlocalization,repairingandtesting.Amongthesesteps,faultlocalizationhasbeenconsideredthemostimportantstepoffaultmanagement.Therefore,wefocusonthestudyoffaultlocalizationandproposedanapproachbasedonAntColonyalgorithmtofaultloca-lizationincomputernetworks.Wealsoevaluatetheproposedapproachbysimulations,andshowthatouralgorithmissuperiortotheotherfaultlocalizationalgorithms.
Withtherisingofvariousnetworkapplications,usersdemandbetterqualityofservice.Oneofthemostcrucialissuesistomaintainthenetworkavailabilityandreliability.Faultsareines-capableincommunicationnetworks,theirimmediatedetectionandisolationisnecessary.Itisnoteasytodowithnetworkfaultsonlybynetworkoperatorsinshorttimeduetothecomplexityofthenetwork.Toshortenthefaultlocalizationresponsetimeandreleasenetworkoperators,automatedfaultmanagementisanacceptablegoalofNetworkManagementSystem(NMS)imple-mentationforlarge-scalenetworks.
Theprocessoffaultmanagementincomputernetworkscom-prisesfoursteps,whicharefaultdetection,faultlocalization,repairingandtesting.Inthefirststep,theadministratorofthesystemshoulddetectthesituationofallthecomponents(faultyornotfaulty).Whenerrors(oralarms)happeninthenetwork,theadministratorshouldreportthemassoonaspossible.Faultlocalizationissearchingtofindexactlytheplaceoffaults.Inthethirdstep,oncethefaultislocated,administratorsmodifythesourcetoerasethefault.Theymustalsoverifywhethertherepairworksbyexecutingthefailedtestcases.Afterthefaultycomponentshasbeenrepaired,thenthesemustre-executethepreviouslypassedtestcasestoensurethatthemodificationdoesnotbreakotherfunctionalities.Inthispaperweintroduceanalgorithmthatusesactiveandpassivemeasurementsforfaultlocalizationinacomputernetworksbyantcolonyalgorithm.Basedonproposedapproach,theantcolonyalgorithmisemployedforextractingthebestcomponentthatshouldbefirsttested,withanaimtominimizethetestingcostofallthefaultycomponentsinanetwork.Theperformanceoftheproposedapproachisevaluatedthroughextensivesimulationindifferentscalesofnetwork.
Faultlocalizationisanimportantpartofnetworkfaultman-agement.Inrecentresearches,alotoffaultlocalizationalgorithmshavebeenproposed.Thesealgorithms,canbepassiveoractivealgorithms.Activealgorithmsimposeexcessivemanagementtraffic,whereaspassivealgorithmsoftenignoreuncertaintyinherentinnetworkalarms,leadingtounreliablefaultidentifica-tionperformance.In,LIetal(2005)hasbeenofferedamobileagent-basedalgorithm.Thehierarchicalstructureprovidedbytheinternetmodelisusedasafaultpropagationmodelandusedasaneventcorrelationapproach.Shuoetal(2009)proposedanovelalgorithmtolocalizefaultynodesinanetworks.Afterthenodesinanetworkdetectanaturalevent,proposedalgorithmranksthenodesbasedontheirfaultprobabilityaswellastheirphysicallocationfromtheevent.Anodeisconsideredfaultyifthereisasignificantmismatchbetweenthetransmitteddatafromthenodeandreceiveddatafromserver.Ramanderetal(2012)proposedamethodtofindthefaultycomponentsinacomputernetworkandprovidesthealternativepathandalsodeterminethenumberoffaultynode.ThismethodlocalizedfaultynodesbytheFloyd–Warshallalgorithm.AnasandTaskin(2009)proposedtwonewalgorithmstolocalizefaultynodesatdifferentlevelsinanetwork.Intheproposedalgorithm,thenet-workisdividedintozonesthatarehavingamasterforzones.Whenafaultoccurs,themastersarechecked,tested.Afterthat,thenodesinthesuspectedzonearetested.Thisfaultmodelassumescommunication,processingfaultscausedbyhardwarefailuresinanode.Batsakisetal(2005)proposedaprobe-baseselectionheuristicalgorithmformonitoringnetworkcomponentsbasedonagreedyalgorithmforchoosingaprobethatcoversthemaximumnumberofuncoverednetworkcomponents.Ozmutluetal(2003)proposedanapproachforfaultlocalizationusingend-to-enddelaydataandcommonoccurrencesofhighdelaysonpathsandthenselectingasubsetofpathstoprobetoidentifyeachcomponentuniquely.Garshasbietal(2014)haveproposedapassiveheuristicalgorithmusingend-to-enddataforfaultlocali-zationincomputernetworks.Garshasbietal(2014)proposedanend-to-endschemeforfaultdiagnosisbasedonheuristicalgo-rithmthatusestheembeddedinformationretrievedfromdis-seminateddataoverthenetworktodetectfaultycomponentsinthecomputernetwork.Maitreyaetal(2006)usedactivelyprobingtopresentanapproachtodeveloptoolsforper-formingfaultlocalization.Theydiscussvariousdesignissuesinvolvedandproposeanarchitectureforbuildingsuchatoolandproposedanalgorithmforprobesetchoiceforproblemdetectionandpresentsimulationresultstoindicateitseffectiveness.Bingetal(2012)proposedanewapproachthatcarefullymixesactiveandpassivetechniquestolocalizefaultycomponentsincomputernetworks.Patriketal(2007)introducedanend-to-endheuristicapproachforsearchfaultincomputernetworks.Theyproposedaproblemwithanoptimizationgoalofminimizingthecostoffaultlocalization.Rostetal(2006)designedahealthfaultmonitoringsystemthatdeliversstatesummariesanddetectsfaultycomponents.Nguyenetal(2007)proposedlossylinkinferenceapproachesthatuseuncorrelatedpacketsandtakeaccountofthedynamicnetworktopologiesto.HsinandLiuproposedadistributedmonitoringmodelthateachnodemonitorsitsneighborsbyperiodicallysendingthemprobes.Rama-nathanetal.proposedatoolfordebuggingfailuresinanetworks.Sympathycarefullychoosesmetricsthatenableefficientfailurelocalizationandincludesanalgorithmthatanalyzesrootcauses.Ntalampirasetal(2015)proposedaschemebaseonprobabilisticmodelingforanalyzingdestructiveeventsrevealingindependentcriticalsubstructure.TheproposedschemeisbasedondesigningtherelationshipbetweendatastreamscomingfromtwonetworknodesbyaHiddenMarkovmodel(HMM)learnedontheparametersoflineartime-invariabledynamicsystemswhichapproximationtherelationshipsavailablebetweenthespecificnodesovercontinuoustimewindows.Caietal(2014)pro-posedanapproachbaseonmulti-sourceinformationfusionbyBayesiannetwork.TheBayesiannetworksbasedonsensordataandapperceivedinformationofhumanbeingarefixed,respec-tively.EachBayesiannetworkcontainedtwolayers:
faultlayerandfaultsymptomlayer.TheentirefaultdiagnosismodelisfixedbycombiningthetwoproposedBayesiannetworks.Baraldietal(2015)proposedacomparisonofthreedata-drivensignalrecon-structionmethodsforfaultdiagnosisbasedonthedifferencebetweenthesignalobservationsandthereconstructionofthesignalinnormaloperatingsituation.JamaliandGarshasbi(2016)proposedanend-to-endschemebypassivemeasurementsandemploythegeneticalgorithmforfaultlocalizationincom-puternetworks.Zhang(2015)expandsthedynamicuncertaincausalitygraphmethodologytodealwithnegativefeedbacks,whichisoneofthehardestproblemsinfaultdiagnosis.Twoschemesareintroduced.Oneisnotbasedoncausalitypropagationacrosstimeslices.Anotherisbasedoncausalitypropagationthroughtimeslices.
Abovemethods,however,canleadtoalargenumberoffalsepositivesandfalsenegativesincertainstatus.Mostofthesemethodsareexpandedforwirednetworksandcan'tbeapplieddirectlytowirelessnetworksespeciallywirelesssensornetworks.Thisisbecausemostofthesemethodstrustondependentpacketsandrequirestatictopology.Indynamictopology,however,end-to-enddataarenotcorrelatedandtopologymaychangeovertime.Proposedmethodinthispaperdiffersfromexistingmethodsonnetworksfaultlocalizationinthatitcarefullycombinesactivemeasurementsandpassivemeasurements.Proposedmethodusesactivemeasurementsdetectfaultypaths,whichinturntriggerpassivemeasurementsandroot-causeanalysis.Inourmethod,thefirstantcolonyalgorithmbysendingants(activemeasurements)acquirespath'sinformationandthencanidentifyfaultynodes(passivemeasurements).
通过蚁群算法来定位处理计算机网络上的故障
摘要:
随着计算机网络规模的持续扩大和复杂,有效的网络管理将变得更加重要,更具挑战性。
计算机网络应用程序存在各种各样的软件或硬件故障。
这些缺点是昂贵的调试和部署的网络的重要原因。
一般来说,计算机网络故障管理包括四个步骤:
故障检测、故障定位、修理和测试。
在这些步骤中,故障定位一直被视为故障管理的最重要的一步。
因此,我们关注的是故障定位的研究,并提出了一种基于蚁群算法的方法在计算机网络故障定位。
我们用评估建议的方法模拟,结果表明我们的算法优于其他故障定位算法。
引言:
随着计算机应用的增多用户需要更好的服务,在这之中最重要的是维持网络的可接受性和可信赖性。
故障是在日常网络工作中所不可避免的,所以及时定位和处理故障是非常有必要的,短时间内很难去定位故障,这是由于网络的复杂性。
如何缩短定位的时间显得尤其重要,自动故障管理是网络管理系统最想完成的一项工作,尤其是在大规模的网络管理中。
计算机网络故障管理包括四个步骤:
故障检测、故障定位、修理和测试.在第一个步骤中计算机管理者必须检查所有的硬件和软件(不管他有没有错误都得检查),当错误或者警告在计算机网络上发生时,计算机管理者要尽可能的去记录它。
故障定位是精确到找到故障的位置。
在第二步中,一旦故障被定位到了,管理者会修改资源管理区清除故障,不管这个修复过程是不是成功,都会修改资源管理区。
在故障部位被修复了以后,他们就要再次检测并测试来保证这次修改不会对其他的硬件造成伤害或损害,在这篇文章中,我们引进了一种新的算法,这种算法可是让用户主动或者被动的去定位故障,尤其是在互联网上的故障。
这是一种蚁群算法,蚁群算法是用于结合其他方法来使用的,首先先