ndn-np_proposal

FIA-NP:CollaborativeResearch:NamedDataNetworking

NextPhase(NDN-NP)

August8,2014

Contents1Vision:ANewNarrowWaist2Architecture

2.1Names.................................................

2.2Data-CentricSecurity........................................

2.3AdaptiveRoutingandForwarding..................................

2.4In-NetworkStorage..........................................

2.5FromTransporttoDistributedSynchronization...........................3ProgressTowardVision:2010-2013

3.1NDNApplicationDevelopment...................................

3.2RoutingandForwardingStrategy..................................

3.3ScalableForwarding.........................................

3.4NetworkExperimentation:Testbed,Simulation,andDemonstration...............

3.5SecurityandPrivacy.........................................

3.6CommunityDevelopmentandOutreach..............................4NetworkEnvironments&Applications

4.1HealthIT:OpenmHealth......................................

4.1.1RelatedPriorWork......................................

4.1.2ResearchApproach.....................................

4.2CyberphysicalSystems:EnterpriseBuildingAutomation&Management............

4.2.1RelatedPriorWork......................................

4.2.2ResearchPlan........................................

4.3MobileMultimediaApplicationCluster...............................5ResearchAgendaforNextPhase

5.1Applications..............................................

5.2SecurityandTrustworthiness....................................

5.2.1TrustandSecurityBuildingBlocks.............................

5.2.2Environment-Speci cSecurity...............................

5.2.3AddressingFutureSecurityChallenges..........................

5.3RoutingandForwardingStrategy..................................

5.4ScalableForwarding.........................................

5.4.1CoreForwardingNodeDesign,Algorithms&Datastructures..............

5.4.2Closely-coupledSubsystems................................

5.5LibraryandToolDevelopmentforApplication-DrivenResearch.................

5.6SocialandEconomicImpacts....................................

1234455557899101010111112131314151517171919192121222223

6EvaluationPlan

7EducationandOutreach

8BroaderImpacts242425

FIA-NP:CollaborativeResearch:NamedDataNetworkingNextPhase(NDN-NP)PROJECTSUMMARY

NamedDataNetworking(NDN)isaFutureInternetArchitectureinspiredbyyearsofempiricalresearchintonetworkusageandagrowingawarenessofpersistentlyunsolvedproblemsofthecurrentInternet(IP)architecture.ItscentralpremiseisthattheInternetisprimarilybeingusedasaninformationdistributionnetwork,ausethatisnotagoodmatchforIP,andthatthefutureInternet’s“thinwaist”shouldbebasedonnameddataratherthannumericallyaddressedhosts.

ThisproposalcontinuesresearchonNDNstartedin2010underNSF’sFIAprogram.Itappliestheprojectteam’sincreasinglysophisticatedunderstandingofNDN’sopportunitiesandchallengestotwona-tionalpriorities–HealthITandCyberphysicalSystems–tofurthertheevolutionofthearchitectureintheexperimental,application-drivenmannerthatprovedsuccessfulinthe rstthreeyears.Inparticular,ourre-searchagendaisorganizedtotranslateimportantresultsinarchitectureandsecurityintolibrarycodethatguidesdevelopmentfortheseenvironmentsandotherkeyapplicationstowardnativeNDNdesigns.Itsimul-taneouslycontinuesfundamentalresearchintothechallengesofglobalscalabilityandbroadopportunitiesforarchitecturalinnovationopenedupby“simply”routingandforwardingdatabasedonnames.

Ourresearchagendaincludes:(1)Applicationdesign,exploringnamingandapplicationdesignpat-terns,supportforrendezvous,discoveryandbootstrapping,theroleanddesignofin-networkstorage,anduseofnewdatasynchronizationprimitives;(2)Securityandtrustworthiness,providingbasicbuildingblocksofkeymanagement,trustmanagement,andencryption-basedaccesscontrolforthenewnetwork,aswellasanticipatingandmitigatingfuturesecuritychallengesfacedinbroaddeployment;(3)Routingandforwardingstrategy,developingandevaluatingpath-vector,link-state,andhyperbolicoptionsforinter-domainrouting,creatingoverallapproachestoroutingsecurityandtrust,aswellasdesigning exibleforwardingandmobilitysupport;(4)Scalableforwarding,aimingtosupportreal-worlddeployment,eval-uationandadoptionviaanoperational,scalableforwardingplatform;(5)Libraryandtooldevelopment,developingreferenceimplementationsforclientAPIs,trustandsecurity,andnewnetworkprimitivesbasedontheteam’sfundamentalresults,aswellassupportinginternalprototypedevelopmentandexternalcom-munityefforts;(6)Socialandeconomicimpacts,consideringthespeci cquestionsfacedinournetworkenvironmentsaswellasbroaderquestionsthatariseinconsideringa“WorldonNDN.”

WechooseMobileHealthandEnterpriseBuildingAutomationandManagementSystemsasspeci cinstancesofHealthITandCyberphysicalSystemstovalidatethearchitectureaswellasdrivenewresearch.DomainexpertsfortheformerwillbetheOpenmHealthteam,anon-pro tpatient-centricecosystemformHealth,ledbyDeborahEstrin(Cornell)andIdaSim(UCSF).Forthelatter,ourexpertswillbeUCLAFacilitiesManagement,operatorsofthesecondlargestSiemensbuildingmonitoringsystemontheWestCoast.ToguideourresearchonthesecuritydimensionsoftheseimportantenvironmentsandtheNDNarchitecturemoregenerally,wehaveconvenedaSecurityAdvisoryCouncil(NDN-SAC)tocomplementourownsecurityandtrusteffort.

IntellectualMeritTheNDNarchitecturebuildsonlessonslearnedfromthesuccessoftheIParchitec-ture,preservingprinciplesofthethinwaist,hierarchicalnames,andtheend-to-endprinciple.Thedesignre ectsarecognitionofthemajorshiftintheapplicationscommunicationmodel:fromthe“where”(i.e.,thehost/location)tothe“what”(i.e.,thecontent).Architectingacommunicationsinfrastructurearoundthisshiftcanradicallysimplifyapplicationdesignstoallowapplicationstocommunicatedirectlyusingthenameofthecontenttheydesireandleavetothenetworkto gureouthowandwheretoretrieveit.NDNalsorec-ognizesthatthebiggestweaknessinthecurrentInternetarchitectureislackofsecurity,andincorporatesafundamentalbuildingblocktoimprovesecuritybyrequiringthatallcontentbecryptographicallysigned.BroaderImpactsThesuccessofnewarchitecturesrequiresbroadcommunityinvolvementanduptake.NDNhasbuiltsigni cantmomentumthroughcommitmenttoanopensourcemodelthathasspurredsub-stantialresearchactivityinbotharchitectureandcurrentimplementation.Projectmembersareofteninvitedtopresentat“futureInternet”meetingsaroundtheworld,andwehaveperformedtwohigh-visibilitydemosofNDN’sabilitytohandlelargescaledistribution.IndustryisalsoshowingincreasinginterestinNDN.Fi-nally,NDNhasalsohadasigni cantimpactonourstudents,yieldingseveralcurrentPh.D.thesesonNDNtopics,industryinternshipsinvolvingNDNresearch,andgraduateandundergraduatecurriculummaterialthatofferacomprehensivealternativetoIPtostimulatediscussionofwhatnetworkarchitecturedesignreallymeans.

PROJECTDESCRIPTION

1Vision:ANewNarrowWaist

In2010weproposedtoexploreanewvisionofcommunicationnetworks,inspiredbyyearsofempiricalresearchintonetworkusage,andagrowingawarenessofpersistentlyunsolvedproblemsofthecurrentInternet(IP)architecture.WhileTCP/IPwasauniqueandground-breakingsolution,itwassolvingaprob-lemofthetelephonyworld:apoint-to-pointconversationbetweentwocommunicationendpoints.Theworldhaschangeddramaticallysincethen,drivenbythespectacularsuccessoftheTCP/IParchitecture.Growthine-commerce,digitalmedia,socialnetworking,andsmartphoneapplicationshasresultedintheInter-netprimarilybeingusedasaninformationdistributionnetwork,ausethatisnotagoodmatchtothearchitecture.ManycomplexitiesandkludgeshavebeenintroducedtomitigatethisinherentmisalignmentbetweentheInternetarchitectureanditsprimaryusetoday.

TheconversationalnatureofIPisembodiedinitsdatagramformat:IPdatagramsidentifycommunica-tionendpoints(theIPdestinationandsourceaddresses).WeproposedtheNamedDataNetwork(NDN)architecture,aimingtogeneralizetoday’sInternetarchitecturebyremovingthisrestriction:namesinanNDNdatagramarehierarchicallystructuredbutotherwisearbitraryidenti ers.ThenameinanNDNdata-gramcanidentifyanything,includingacommunicationendpoint,asTCP/IPdoestoday,orachunkofdatainaconversation,astheTCP/IPtransportsignatureplussequencenumberdoestoday.ButNDNnamescanalsoidentifyachunkofamovieorabook,acommandtoturnonsomelights,etc.Thissimplechangetothehourglassmodel,allowingthethinwaisttousedatanamesinsteadofIPaddressesfordatadelivery,makesdataratherthanitscontainersa rst-classcitizenintheInternetarchitecture.Thisconceptu-allysimplechangeallowsNDNnetworkstousealmostalloftheInternet’swell-understoodandwell-testedengineeringpropertiestoef cientlysolvenotonlycommunicationproblemsbutalsodigitaldistributionandcontrolproblems,andtoeffectivelyaddresssomeoftheInternet’smostpressingproblemsinsecurity,scalability,mobility,andfacilitationofapplicationsdevelopment.Inparticular,

NDNenablesapplicationstosecuredataend-to-end;thenameprovidesessentialcontextforsecurity,with ne-grainedsigningandveri cation,limitingthedatasecurityperimetertothecontextofasingleapplication.

NDNnetworksareloop-freebydesign,whichmeansanynodecanfreelyuseany/allofitsconnectivitytosolicitordistributedata.TogetherwithNDN’ssupportforin-networkstorage,thiscapabilityfacilitatesretrievingdatafromnearbylocations,andreducesthecontrolasymmetriesthatgivetoday’sdominantprovidersdisproportionatecontroloverroutesandthusoversmaller,localproviders.

BecauseNDNretrievesdatabynames,itcircumventstheneedforIPaddressesandtheassociatedissueswiththeirscarcity.Mobility,whichrequireschangingaddressesinIP,isseamlessinNDNsincedatanamesremainthesame.

Today’sapplicationsaretypicallywrittenintermsofwhatinformationtheywantratherthanwhereit’slocated,usingapplication-speci cmiddlewaretomapfromapplicationtonetworklayer.NDNenablesapplicationstodirectlyaddresswhattheywant,simplifyingapplicationdevelopmentandenablingfunda-mentallynewtypesofapplications.

Wearepursuinganapplication-drivenapproachtoarchitectureresearchanddevelopment,andhavedeployeda12-sitetestbed(anoverlayonthecurrentInternet)onwhichtoiterativelydesignthem.Oursetofprototypeapplicationsworkintheirlocalenvironment(e.g.buildingcontrol, lesharing)andprovideimmediatebene ts.Theyalsoworkacrossthetestbed,andwehavedemonstratedseveralcommunicationsanddatadistributionapplicationsonaglobaldemonstrationenvironment(describedinSection3.4).

Ourapplication-driveninvestigationhasenabledustobegindescribinggeneralprinciplesandguide-linesforapplicationdesignsinNDNnetworks,andtotranslatetheseprinciplesandguidelinesintonamingconventionsimplementedinsystemlibraries.Libraryimplementationoperationalizeswhatwehavelearnedinaformthatsupportsconsistentreuse,simplifyfutureapplicationdevelopmentandaccelerateprogress,andpromotescommunityexperimentationandcontribution.BroadR&Dcommunityengagementisneces-sarytothesuccessofanynewarchitecture;weareencouragedbythemanyinformation-centricnetworkingworkshopsthathaveproliferatedoverthelast3years.

Inthisproposalwe rstpresentareviewofthecurrentNDNarchitectureinSection2.Section3sum-marizesourprogressinthelastthreeyears.Section4describesthetwonetworkenvironmentswehave

内容需要下载文档才能查看 内容需要下载文档才能查看 内容需要下载文档才能查看 内容需要下载文档才能查看 内容需要下载文档才能查看 内容需要下载文档才能查看

Every node

Figure1:InternetandNDNHourglassArchitectures

selectedtotestandevaluateourarchitectureinthenextphaseoftheFIAproject.InSection5weprovidedetailsonourresearchagenda,includingapplication,routingandforwardingstrategies,scalableforward-ing,softwaredevelopment,andsocialandeconomicimpacts.Section6describesourapproachtoevalua-tionofthearchitecture,andSection7explainsourplanforeducationandoutreach.WesummarizebroaderimpactcontributionsinSection8.OurmanagementandcollaborationplanandintellectualpropertypolicystatementsareinseparatedocumentspertheFIA-NPsolicitationrequirements.

2Architecture

ThedesignprinciplesofNDNre ectourunderstandingofthestrengthsandlimitationsofthecurrentInternetarchitecture.TheInternet’shourglassarchitecturecentersonauniversalnetworklayer(i.e.,IP)whichimplementstheminimalfunctionalitynecessaryforglobalinterconnectivity.ThisthinwaistenabledtheInternet’sexplosivegrowthbyallowingbothlowerandupperlayertechnologiestoinnovatewithoutunnecessaryconstraints.TheNDNarchitectureretainsthisextraordinarilysuccessfulhourglass-shapedarchitecture,buttransformsthethinwaisttofocusondatadirectlyratherthanitslocation.Morespeci cally,NDNchangesthesemanticofthenetworkservicefromdeliveringapackettoagivendestinationaddresstoretrievingdataidenti edbyagivenname(Figure1).NDN’sdesignisalsoguidedbythefollowingprinciples. Securitymustbebuiltintothearchitecture.SecurityinthecurrentInternetarchitectureisanafterthought,

notmeetingthedemandsoftoday’sdiverseenvironment.NDNprovidesafundamentalsecuritybuildingblockrightatthethinwaistbysigningallnameddata.

Theend-to-endprinciple[61]underlyingtheTCP/IParchitectureenableddevelopmentofrobustappli-

cationsinthefaceofunexpectedfailures.NDNexpandsthisprinciplebysecuringdataend-to-end.

Networktraf cmustbeself-regulating.Flow-balanceddatadeliveryisessentialtothestabilityoflarge

systems.UnlikeIP’sopen-looppacketdelivery,NDNdesigns ow-balanceintothethinwaist.

Thearchitectureshouldfacilitateuserchoiceandcompetitionwherepossible.Althoughnotconsidered

intheoriginalInternetdesign,globaldeploymenthastaughtusthat“architectureisnotneutral.”NDNmakesaconsciousefforttoempowerendusersandfacilitatecompetition[17].

NDNcommunicationconsistsoftwotypesofpackets:InterestandData(seeFigure2).Bothtypesofpacketscontainanamethatidenti esapieceofdatathatcanbetransmittedinoneDatapacket.ToretrieveData,aconsumerputsthenameofdesireddataintoanInterestpacketandsendsittothenetwork.RoutersusethisnametoforwardtheInteresttowardthedataproducer(s).OnceanInterestreachesanodethathastherequesteddata,thatnodereturnsaDatapacketthatcontainsboththenameandthecontent,andaproducersignaturethatbindsthetwo.NDNcommunicationisalwaysreceiver-drivenandtwo-way.

ConceptuallyeachNDNroutermaintainsthreemajordatastructures:ContentStore,PendingInterestTable(PIT),andForwardingInformationBase(FIB)(Fig.3).TheContentStoreisatemporarycacheofDatapacketsthattherouterhasreceived.BecauseanNDNDatapacketismeaningfulindependentofwhereitcomesfromorisheaded,theroutercancacheittosatisfyfutureInterests.ThePITstoresallIntereststhathavebeenforwardedbutnotsatis edyet.ItrecordstheInterest’sname,incominginterface(s)andoutgoinginterface(s).WhenanodereceivesmultipleInterestswiththesamenamefromdownstream,itneedstosendonlythe rstoneupstreamtowardthedataproducer.TheFIBisaname-pre x-based