可逆bloomfilter及其在長(zhǎng)流識(shí)別應(yīng)用中的研究

1、大連海事大學(xué)碩士學(xué)位論文可逆BloomFilter及其在長(zhǎng)流識(shí)別應(yīng)用中的研究姓名：霍丹丹申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)科學(xué)與技術(shù)指導(dǎo)教師：劉衛(wèi)江20090701英文摘要AbstractIdentifyingelephantflowsisveryusefulfortrafficengineeringschemes，networkoperationandmanagementManymeasurementstudieshaverevealed

2、thatflowstatisticsexhibitstrongheavytailedbehaviorsinvariousnetworksThischaracteristicisoftenreferredtoastheelephantandmicephenomenon；ie，mostflows(miceflows)onlyhaveasmallnumberofpackets，whileaveryfewflows(elephantflows)

3、havealargenumberofpacketsElephantflowscontributealargepotionofthetotaltrafficvolumedespitebeingrelativelyfewinthenumberoffowsKnowingtheinformationofelephantflowscanmasterthetra伍CbehavioraccordinglyInthispaperwestudyandim

4、provetheinvertibleBloomFilterforindentifyingelephantflowsFirstweintroducetheapplicationconditionofBloomFilteranditsvariations，thenputtheaccentontheinvertibleBloomFilterTheinvertibleBloomFilterisalsooneofthesevariationsIt

5、hasbeenproposedasamethodofnetworkmeasurementforafewyearsTheinvertibleBloomFilterconsistsof8hashfunctionswhichselectsomebitsfromoriginalstringsasfunctionvaluesWeobtaintheidentifierandlengthofelephantflowsaccordingthechara

6、cteristicsofhashfunctonsandheavytailedfeatureofflowdistribution。SotheinvertibleBloomFilterdoesnotneedtokeepaflowidentifierforeveryhashtablebodyInaddition，itassignseveryhashfunctionanindependentspacewhichremovestheinterna

7、lconflictionamonghashingAtlast，wejustapplyrecoveralgorithmtosomestringswhosenumbersarebiggerthanthethresholdThatCanreducecalculatedamountgreatlyWedoexperimentwiththetracescollectedfromCERNETTheexperimentshowsthattheident

8、ifiersandlengthofelephantflowsCanbeobtainedaccuratelythroughthealgorithmThoughtheresearchanddevelopmentoftheinvertibleBloomFilterhaveyieldedsomeresults，theinvertibleBloomFilterstillisanewmethodwhichneedsmorewoktoimproveT