版權(quán)說明：本文檔由用戶提供并上傳，收益歸屬內(nèi)容提供方，若內(nèi)容存在侵權(quán)，請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)
文檔簡(jiǎn)介
APrediction-basedFairReplicationAlgorithminStructuredP2PSystemsXianshuZhu,DafangZhang,WenjiaLi,KunHuangPresentedby:XianshuZhuCollegeofComputer&Communication,HunanUniversity,P.R.ChinaAPrediction-basedFairReplic1OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction2IntroductionQueryHotspotStructuredPeer-to-PeerNetworkSummaryofReplicationSchemes
IntroductionQueryHotspot3QueryHotspot
FGIJCDEHBFileQueryHotspot:thenumberofrequestsforpopularobjectsincreasesdramatically,andleadstoconsequentdroppingqueriesandsevereperformancefailures.QueryHotspotQueryHotspotFGIJCDEHBFileQue4StructuredP2PNetworkAdvantage：
-Scalability-EfficientSearchingDisadvantage：TheImplementationofStructuredP2PNetworkAssumesthatAllDataItemsareoftheSamePopularity.NoMechanismCanHandleHotspotProblem
StructuredP2PNetworkAdvantag5ReplicationSchemesBasicIdea：-DistributeReplicasofthePopularDataItemstoVariousLight-loadedNodes-Fairly
DistributeLoadontoEachNode.WhenApplyReplicationTechnique:－ReplicaCreation:Time,Number,Location－ReplicaUtilization
ReplicationSchemesBasicIdea：6ReplicationSchemesClassificationAccordingtoReplicaLocation:-PathReplication-OwnerReplication-RandomReplicationABCDEFFileFileFileFileFileFileHighReplicationOverheadReplicationSchemesClassificat7ReplicationSchemesABCDEFFileA1.NewQueryHotspot2.LowReplicationSpeedClassificationAccordingtoReplicaLocation:-PathReplication
-OwnerReplication:GopalakrishnanproposedLAR-RandomReplicationFileBFileDFileDFileBFileAReplicationSchemesABCDEFFile8ReplicationSchemesABCDEFFileFileFileFileClassificationAccordingtoReplicaLocation:-PathReplication-OwnerReplication
-RandomReplicationReplicationSchemesABCDEFFileF9OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction10ContributionDesignGoals:-DroppedQueriesbyOnlyIntroducingMinimumReplicationOverhead-MinimizetheDrawbacksofLARAlgorithm
(OwnerReplication)Prediction-basedFairReplicationAlgorithm(PFR)thatCanAlmostFairlyDistributeLoadontoEachNode,SoAstoMeettheAboveDesignGoal.
ContributionDesignGoals:11ContributionFairnessGoalofPFR
-AdaptivelyDeterminetheReplicationSpeedandReplicationLocationAccordingtoNode’sPredictedLoadFractionABCDEFGContributionFairnessGoalofP12OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction13Predict(n+1)PFR-AppropriateReplicationTime
TokeeptheSystemPerformanceataHighLevel,PreventiveActionsShouldbeTakenBeforeQueryHotspotReallyHappensPeriodExponentialWeightPredictionAlgorithmPredict(n+1)=Current(n)+PredictDiff(n+1)12nn+1n-1CurrentTimePredictedPossibleTrafficDifferenceBetweennthand(n+1)thInterval
Predict(n+1)PFR-AppropriateR14PeriodExponentialWeightPredictionAlgorithm-OnlyIncursLowComputationOverhead-ApplicabletoOnlinePredictionOurReplicationStrategyisSetBasedonThePredictedload
PFR-AppropriateReplicationTime
PeriodExponentialWeightPred15ReplicationSpeed:ABCDEFFileFileFileFile3/6ReplicationSpeed=(theNumberofNodesChosentoHoldReplicas)/(theNumberofAllNodesthatHaveEncounteredAlongtheQueryPath)PFR-Fairly-decidedReplicationSpeedReplicationSpeed:ABCDEFFileFi16ReplicationLevel:NN/23N/4N/41DON’TcreatereplicasN:TotalNumberofNodesAlongaQueryPathPFR-Fairly-decidedReplicationSpeedReplicationSpeedPredictedLoadFraction(0.5)(0.3)(0.6)(0.7)(0.8)(1)NodeHomogeneityReplicationLevel:NN/23N/4N/4117PFR-Replication&ReplicaUtilizationABCDEFGC:FileF:0.25E:0.15F:0.25E:0.15F:0.25D:0.3C:0.55E:0.15F:0.25D:0.3B:0.3C:0.55E:0.15F:0.25D:0.3A:0.9B:0.3C:0.55E:0.15F:0.25D:0.3RS:N/4=1A:FileA:FileA:FileA:FileA:FileRS:NE:CE:CE:CB,D,E,F:AB,D,E,F:AB,D,E,F:AB,D,E,F:AB,D,E,F:AB,D,E,F:AD:AN=6PFR-Replication&ReplicaUti18OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction19PerformanceEvaluation
HighlymodifiedChordSimulatorfromMITandLARImplementationCode：SystemSize1000TheTimeEachNetworkhoptakes25msNumberofdata32767Averagesystemload25%Nodecapacity10persecNumberofQueriesGenerateperSec500Node’squeuelength32Predictioninterval1sPerformanceEvaluationHighly20PerformanceEvaluationNumberofQueriesDroppedOverTime
28%90%oftheinputqueriesaredirectedto1
itemLARPFRPerformanceEvaluationNumbero21PerformanceEvaluationTotalNumberofDocumentsReplicatedLARPFRPerformanceEvaluationTotalNu22PerformanceEvaluationTotalNumberofFingerTablesReplicatedLARPFRPerformanceEvaluationTotalNu23PerformanceEvaluationTotalNumberofReplicaLocationHintsCreatedPFRLARPerformanceEvaluationTotalNu24OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction25ConclusionPrediction-basedFairReplicationAlgorithmCanConductFairReplicationthrough:-AppropriateReplicationTime-Fairly-decidedReplicationSpeed-Fairly-decidedReplicationLocation-HighReplicaUtilizationRatePerformanceEvaluation:-NotablyDecreasetheNumberofDroppedQueries-LowReplicationOverhead
ConclusionPrediction-basedFai26FutureWorkTakingNodeHeterogeneityintoConsiderationFutureWorkTakingNodeHeterog27Thankyou!Thankyou!28APrediction-basedFairReplicationAlgorithminStructuredP2PSystemsXianshuZhu,DafangZhang,WenjiaLi,KunHuangPresentedby:XianshuZhuCollegeofComputer&Communication,HunanUniversity,P.R.ChinaAPrediction-basedFairReplic29OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction30IntroductionQueryHotspotStructuredPeer-to-PeerNetworkSummaryofReplicationSchemes
IntroductionQueryHotspot31QueryHotspot
FGIJCDEHBFileQueryHotspot:thenumberofrequestsforpopularobjectsincreasesdramatically,andleadstoconsequentdroppingqueriesandsevereperformancefailures.QueryHotspotQueryHotspotFGIJCDEHBFileQue32StructuredP2PNetworkAdvantage：
-Scalability-EfficientSearchingDisadvantage：TheImplementationofStructuredP2PNetworkAssumesthatAllDataItemsareoftheSamePopularity.NoMechanismCanHandleHotspotProblem
StructuredP2PNetworkAdvantag33ReplicationSchemesBasicIdea：-DistributeReplicasofthePopularDataItemstoVariousLight-loadedNodes-Fairly
DistributeLoadontoEachNode.WhenApplyReplicationTechnique:－ReplicaCreation:Time,Number,Location－ReplicaUtilization
ReplicationSchemesBasicIdea：34ReplicationSchemesClassificationAccordingtoReplicaLocation:-PathReplication-OwnerReplication-RandomReplicationABCDEFFileFileFileFileFileFileHighReplicationOverheadReplicationSchemesClassificat35ReplicationSchemesABCDEFFileA1.NewQueryHotspot2.LowReplicationSpeedClassificationAccordingtoReplicaLocation:-PathReplication
-OwnerReplication:GopalakrishnanproposedLAR-RandomReplicationFileBFileDFileDFileBFileAReplicationSchemesABCDEFFile36ReplicationSchemesABCDEFFileFileFileFileClassificationAccordingtoReplicaLocation:-PathReplication-OwnerReplication
-RandomReplicationReplicationSchemesABCDEFFileF37OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction38ContributionDesignGoals:-DroppedQueriesbyOnlyIntroducingMinimumReplicationOverhead-MinimizetheDrawbacksofLARAlgorithm
(OwnerReplication)Prediction-basedFairReplicationAlgorithm(PFR)thatCanAlmostFairlyDistributeLoadontoEachNode,SoAstoMeettheAboveDesignGoal.
ContributionDesignGoals:39ContributionFairnessGoalofPFR
-AdaptivelyDeterminetheReplicationSpeedandReplicationLocationAccordingtoNode’sPredictedLoadFractionABCDEFGContributionFairnessGoalofP40OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction41Predict(n+1)PFR-AppropriateReplicationTime
TokeeptheSystemPerformanceataHighLevel,PreventiveActionsShouldbeTakenBeforeQueryHotspotReallyHappensPeriodExponentialWeightPredictionAlgorithmPredict(n+1)=Current(n)+PredictDiff(n+1)12nn+1n-1CurrentTimePredictedPossibleTrafficDifferenceBetweennthand(n+1)thInterval
Predict(n+1)PFR-AppropriateR42PeriodExponentialWeightPredictionAlgorithm-OnlyIncursLowComputationOverhead-ApplicabletoOnlinePredictionOurReplicationStrategyisSetBasedonThePredictedload
PFR-AppropriateReplicationTime
PeriodExponentialWeightPred43ReplicationSpeed:ABCDEFFileFileFileFile3/6ReplicationSpeed=(theNumberofNodesChosentoHoldReplicas)/(theNumberofAllNodesthatHaveEncounteredAlongtheQueryPath)PFR-Fairly-decidedReplicationSpeedReplicationSpeed:ABCDEFFileFi44ReplicationLevel:NN/23N/4N/41DON’TcreatereplicasN:TotalNumberofNodesAlongaQueryPathPFR-Fairly-decidedReplicationSpeedReplicationSpeedPredictedLoadFraction(0.5)(0.3)(0.6)(0.7)(0.8)(1)NodeHomogeneityReplicationLevel:NN/23N/4N/4145PFR-Replication&ReplicaUtilizationABCDEFGC:FileF:0.25E:0.15F:0.25E:0.15F:0.25D:0.3C:0.55E:0.15F:0.25D:0.3B:0.3C:0.55E:0.15F:0.25D:0.3A:0.9B:0.3C:0.55E:0.15F:0.25D:0.3RS:N/4=1A:FileA:FileA:FileA:FileA:FileRS:NE:CE:CE:CB,D,E,F:AB,D,E,F:AB,D,E,F:AB,D,E,F:AB,D,E,F:AB,D,E,F:AD:AN=6PFR-Replication&ReplicaUti46OutlineIntroductionContributionPFR(Prediction-basedFairReplication)PerformanceEvaluationConclusionandFutureWorkOutlineIntroduction47PerformanceEvaluation
HighlymodifiedChordSimulatorfromMITandLARImplementationCode：SystemSize1000TheTimeEachNetworkhoptakes25msNumberofdata32767Averagesystemload25%Nodecapacity10persecNumberofQueriesGeneratepe
溫馨提示
- 1. 本站所有資源如無特殊說明，都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等，如果需要附件，請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
- 3. 本站RAR壓縮包中若帶圖紙，網(wǎng)頁內(nèi)容里面會(huì)有圖紙預(yù)覽，若沒有圖紙預(yù)覽就沒有圖紙。
- 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫網(wǎng)僅提供信息存儲(chǔ)空間，僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理，對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯，并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
- 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容，請(qǐng)與我們聯(lián)系，我們立即糾正。
- 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。
最新文檔
- 2025年度個(gè)人住房抵押貸款還款管理協(xié)議4篇
- 2025版攝影棚租賃合同涵蓋廣告、商業(yè)拍攝6篇
- 2025年度水利工程個(gè)人承包協(xié)議書2篇
- 2025版地質(zhì)勘探打井合同范本3篇
- 二零二五年度車輛運(yùn)輸服務(wù)與貨物跟蹤系統(tǒng)合作協(xié)議2篇
- 2025年度魚塘承包權(quán)抵押貸款服務(wù)合同4篇
- 二零二五年度橙子出口歐盟認(rèn)證采購合同3篇
- 2025年度個(gè)人房屋維修欠款合同模板4篇
- 二零二五年度畜牧養(yǎng)殖生物安全防控體系建設(shè)合同4篇
- 2025年度個(gè)人房屋買賣合同履行監(jiān)督及保障協(xié)議2篇
- 2025至2030年中國減肥肽數(shù)據(jù)監(jiān)測(cè)研究報(bào)告
- 2024內(nèi)蒙古公務(wù)員省直行測(cè)、行政執(zhí)法、省考行測(cè)考試真題（5套）
- 山東省濱州市濱城區(qū)2024-2025學(xué)年九年級(jí)上學(xué)期期末考試化學(xué)試題
- 貨運(yùn)企業(yè)2025年度安全檢查計(jì)劃
- 2025年焊工安全生產(chǎn)操作規(guī)程（2篇）
- 《事故快速處理協(xié)議書》電子版
- 參考新醫(yī)大-中央財(cái)政支持地方高校發(fā)展專項(xiàng)資金建設(shè)規(guī)
- 山東省房屋市政工程安全監(jiān)督機(jī)構(gòu)人員業(yè)務(wù)能力考試題庫-上（單選題）
- 松下-GF2-相機(jī)說明書
- 產(chǎn)教融合背景下“一體兩翼三融合五重點(diǎn)”創(chuàng)新創(chuàng)業(yè)人才培養(yǎng)機(jī)制研究
- 第21課 活動(dòng)課 從考古發(fā)現(xiàn)看中華文明的起源 教學(xué)課件
評(píng)論
0/150
提交評(píng)論