欢迎来到天天文库
浏览记录
ID:33583012
大小:537.33 KB
页数:41页
时间:2019-02-27
《advancement proposal storage embedded networks (sen) and adaptive caching using multiple ex》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、AdvancementProposalStorageEmbeddedNetworks(SEN)andAdaptiveCachingusingMultipleExperts(ACME)IsmailAriari@cs.ucsc.eduUCSC-CRL-03-01June2,2003DepartmentofComputerScienceUniversityofCalifornia,SantaCruz0ABSTRACTThegapbetweenCPUspeedsandthespeedofthetechnologiesp
2、rovidingthedataisincreasing.Thiscausestheperformanceofprocessestobelimitedbytheperformanceofthestoragedevices,thenetworksandthebuses.Furthermore,thenumberofCPUsthatsharethesedataaccessresourcesisgrowingexponentially.Caching,prefetchingandparallelismaresomeof
3、thetechniquesusedtodaytocopewithI/Olatencyandsystemscalabilitytosupportmoreusers.Thispaperdescribesthetwomajorcontributionsofourongoingresearchondistributeddataaccess.ThefirstcontributionisthedesignoftheStorageEmbeddedNetworks(SEN)architecturethataimstoimprov
4、euserresponsetimesandscalabilityontheInternetbybetterdistributionofcaches.SENarchitectureiscomposedoftrustedroutersembeddedwithvolatileandnon–volatilestoragethatsnoopbypassingobjectsforcaching.Requestsarecheckedbyeveryhop,thusensuringthetransmissionoftheclos
5、estcopyonthedatapathandloadreductionattheupstream.Thetwomaincontroloverheadsofotherarchitectures,connectionestablishmentandcontinuouscachecommunications,donotexistinSEN.Thesecondcontributionisthedesignofadaptivecachingschemesusingmultipleex-perts,calledACME,
6、thatmanagetheSENcachesandfurtherimprovethehitratesoverstaticcachingtechniques.Machinelearningalgorithmsareusedtorateandselectthecur-rentbestpoliciesormixturesofpoliciesviaweightupdatesbasedontheirrecentsuccess.Eachadaptivecachenodecantuneitselfbasedonthework
7、loaditobserves.Sincenocachedatabasesorsynchronizationmessagesareexchangedforadaptivity,theclusterscomposedofthesenodeswillbeexceedinglyscalableandmanageable.Weproposetoextendourpreliminarydesignsandanalysisintwodirections.ThefirstistocomparetheStorageEmbedded
8、Networks(SEN)withtheexistinghierarchicalanddistributedcacheclustersintermsofuserresponsetimes,networkbandwidthusage,serverloadreductionsandscalability.Forthispartwewillrunlargescalesimulationsov
此文档下载收益归作者所有