资源描述:
《MAsCOT Self-adaptive Opportunistic Offloading吉祥物:自适应机会卸载 支持云计算的智能移动应用程序 在运行时使用概率图形模型》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、201649thHawaiiInternationalConferenceonSystemSciencesMAsCOT:Self-adaptiveOpportunisticOffloadingforCloud-EnabledSmartMobileApplicationswithProbabilisticGraphicalModelsatRuntimeNayyabZiaNaqvi∗,JonasDevlieghere†,DavyPreuveneers∗,andYolandeBerbers∗∗iMinds-DistriNet,KULeuven,BelgiumEmail:∗{n
2、ayyab.naqvi,davy.preuveneers,yolande.berbers}@cs.kuleuven.be†jonas.devlieghere@student.kuleuven.beAbstractitismorerelevantthanevernowwithmorepowerfulmobiledevicesandresourceaswellasdata-intensiveapplicationAlthoughextensiveprogresshasbeenmadeinMobileCloudAugmentation,demands.Itisastocha
3、sticmulti-objectiveoptimizationproblemautomateddecisionsupportonthedevicethatenablestheopportunisticandintelligentuseofcloudresourcesismissing.Furthermore,weneedsolutionswheremanyQuality-of-Service(QoS)trade-offs(computationwithreflectivecapabilitiesthatcanhandleachangingenvironmentandru
4、ntimevs.communication,performancevs.latency,communicationvariability.Tosimplifythedeploymentofsmartmobileapplications,wepresentvs.localityetc.)exist.Thesetrade-offsarecontext-driven,aframeworkwithretrospectivedecisionsupportbasedonreinforcementlearningtocaterforvariousresource-performan
5、cetrade-offs.Wehaveapplicationspecific,anddependontheavailabilityofafastadoptedtheMAPE-K(Monitor-Analyse-Plan-Execute-Knowledge)controlloopinternetconnection,thesizeofthedatatobeprocessed,thearchitectureandrealizedtheloopwithDynamicDecisionNetworkstomanagecloud’sperformanceversusthatofth
6、emobile,andotherfactors.self-adaptationatruntime.OurexperimentsshowthatourframeworkiscapableofintelligentlyinferringappropriatedecisionswithanacceptableRecentoptimizationapproaches[3]offeralimitednumberperformanceoverheadof10millisecondsonmobiledevices.ofobjectivestocaterforresource-per
7、formancetrade-offs.Theoptimizationfactorsvaryfromframeworktoframework.Furthermore,developerscannotinteractnorchangetheKeywordsobjectivesortrade-offs.Additionally,offloadingdecisionsarenaiveratherthanself-adaptive.ThemainchallengeaddressedMobilecloudcomputing,decisionsupport,prob