,Berkeley,,Finland,,XiuwenLiuatFloridaStateUniversityIhavemodifiedthemandaddednewslidesGivingcreditwherecreditisdue:CSCE455/855DistributedOperatingSystemsFaultToleranceChapter8PartIIntroductionFaultToleranceADSshouldbefault-tolerantShouldbeabletocontinuefunctioninginthepresenceoffaultsFaulttoleranceisrelatedtodependabilityDependabilityDependabilityIncludesAvailabilityReliabilitySafetyMaintainabilityAvailability&Reliability(1)Availability:AmeasurementofwhetherasystemisreadytobeusedimmediatelySystemisupandrunningatanygivenmomentReliability:AmeasurementofwhetherasystemcanruncontinuouslywithoutfailureSystemcontinuestofunctionforalongperiodoftimeAvailability&Reliability(2)Asystemgoesdown1ms/%,butisunreliableAsystemthatnevercrashesbutisshutdownforaweekonceeveryyearis100%reliablebutonly98%availableSafety&MaintainabilitySafety:AmeasurementofhowsafefailuresareSystemfails,nothingserioushappensForinstance,highdegreeofsafetyisrequiredforsystemscontrollingnuclearpowerplantsMaintainability:AmeasurementofhoweasyitistorepairasystemAhighlymaintainablesystemmayalsoshowahighdegreeofavailabilityFailurescanbedetectedandrepairedautomatically?Self-healingsystems?FaultsAsystemfailswhenitcannotmeetitspromises(specifications)AnerrorispartofasystemstatethatmayleadtoafailureAfaultisthecauseoftheerrorFault-Tolerance:thesystemcanprovideserviceseveninthepresenceoffaultsFaultscanbe:Transient(appearonceanddisappear)Intermittent(appear-disappear-reappearbehavior)AloosecontactonaconnectorintermittentfaultPermanent(appearandpersistuntilrepaired)FailureModelsTypeoffailureDescriptionCrashfailureAserverhalts,butisworkingcorrectlyuntilithaltsOmissionfailureReceiveomissioningrequestsingmessagesAserver
Fault Tolerance 来自淘豆网www.taodocs.com转载请标明出处.