T-REC-Y.2501-202109-IPDF-E 國際電信聯(lián)盟電信標準分局算力網絡框架與架構標準

InternationalTelecommunicationUnion

Y.2501

(09/2021)

ITU-T

TELECOMMUNICATION

STANDARDIZATIONSECTOR

OFITU

SERIESY:GLOBALINFORMATIONINFRASTRUCTURE,INTERNETPROTOCOLASPECTS,NEXT-GENERATIONNETWORKS,INTERNETOFTHINGSANDSMARTCITIES

NextGenerationNetworks–Computingpowernetworks

Computingpowernetwork–Frameworkandarchitecture

RecommendationITU-TY.2501

ITU-TY-SERIESRECOMMENDATIONS

GLOBALINFORMATIONINFRASTRUCTURE,INTERNETPROTOCOLASPECTS,NEXT-GENERATIONNETWORKS,INTERNETOFTHINGSANDSMARTCITIES

GLOBALINFORMATIONINFRASTRUCTURE

Y.100–Y.199Y.200–Y.299Y.300–Y.399Y.400–Y.499Y.500–Y.599Y.600–Y.699Y.700–Y.799Y.800–Y.899

Y.1000–Y.1099Y.1100–Y.1199Y.1200–Y.1299Y.1300–Y.1399Y.1400–Y.1499Y.1500–Y.1599Y.1600–Y.1699Y.1700–Y.1799Y.1800–Y.1899Y.1900–Y.1999

Y.2000–Y.2099Y.2100–Y.2199Y.2200–Y.2249Y.2250–Y.2299Y.2300–Y.2399Y.2400–Y.2499

General

Services,applicationsandmiddleware

Networkaspects

Interfacesandprotocols

Numbering,addressingandnaming

Operation,administrationandmaintenance

Security

Performances

INTERNETPROTOCOLASPECTS

General

Servicesandapplications

Architecture,access,networkcapabilitiesandresourcemanagement

Transport

Interworking

Qualityofserviceandnetworkperformance

Signalling

Operation,administrationandmaintenance

Charging

IPTVoverNGN

NEXTGENERATIONNETWORKS

Frameworksandfunctionalarchitecturemodels

QualityofServiceandperformance

Serviceaspects:Servicecapabilitiesandservicearchitecture

Serviceaspects:InteroperabilityofservicesandnetworksinNGN

EnhancementstoNGN

Networkmanagement

Y.2500–Y.2599

Computingpowernetworks

Y.2600–Y.2699Y.2700–Y.2799Y.2800–Y.2899Y.2900–Y.2999Y.3000–Y.3499Y.3500–Y.3599Y.3600–Y.3799Y.3800–Y.3999

Y.4000–Y.4049Y.4050–Y.4099Y.4100–Y.4249Y.4250–Y.4399Y.4400–Y.4549Y.4550–Y.4699Y.4700–Y.4799Y.4800–Y.4899Y.4900–Y.4999

Packet-basedNetworks

Security

Generalizedmobility

Carriergradeopenenvironment

FUTURENETWORKS

CLOUDCOMPUTING

BIGDATA

QUANTUMKEYDISTRIBUTIONNETWORKS

INTERNETOFTHINGSANDSMARTCITIESANDCOMMUNITIES

General

Definitionsandterminologies

Requirementsandusecases

Infrastructure,connectivityandnetworks

Frameworks,architecturesandprotocols

Services,applications,computationanddataprocessing

Management,controlandperformance

Identificationandsecurity

Evaluationandassessment

Forfurtherdetails,pleaserefertothelistofITU-TRecommendations.

Rec.ITU-TY.2501(09/2021)i

RecommendationITU-TY.2501

Computingpowernetwork–Frameworkandarchitecture

Summary

RecommendationITU-TY.2501describestheframeworkandarchitectureofthecomputingpowernetwork(CPN).Thisisanewtypeofnetworkthatrealizesoptimizedresourceallocation,bydistributingcomputing,storage,networkandotherresourceinformationofservicenodesthroughanetworkcontrolplane.Itcombinesnetworkcontextanduserrequirementstoprovidetheoptimaldistribution,association,transactionandschedulingofcomputing,storageandnetworkresources.

History

EditionRecommendationApprovalStudyGroupUniqueID*

1.0ITU-TY.25012021-09-1313

11.1002/1000/14768

Keywords

Architecture,computingpowernetwork,framework.

*ToaccesstheRecommendation,typetheURL/intheaddressfieldofyourwebbrowser,followedbytheRecommendation'suniqueID.Forexample,

/11.1002/1000/11

830-en

.

iiRec.ITU-TY.2501(09/2021)

FOREWORD

TheInternationalTelecommunicationUnion(ITU)istheUnitedNationsspecializedagencyinthefieldoftelecommunications,informationandcommunicationtechnologies(ICTs).TheITUTelecommunicationStandardizationSector(ITU-T)isapermanentorganofITU.ITU-Tisresponsibleforstudyingtechnical,operatingandtariffquestionsandissuingRecommendationsonthemwithaviewtostandardizingtelecommunicationsonaworldwidebasis.

TheWorldTelecommunicationStandardizationAssembly(WTSA),whichmeetseveryfouryears,establishesthetopicsforstudybytheITU-Tstudygroupswhich,inturn,produceRecommendationsonthesetopics.

TheapprovalofITU-TRecommendationsiscoveredbytheprocedurelaiddowninWTSAResolution1.

InsomeareasofinformationtechnologywhichfallwithinITU-T'spurview,thenecessarystandardsarepreparedonacollaborativebasiswithISOandIEC.

NOTE

InthisRecommendation,theexpression"Administration"isusedforconcisenesstoindicatebothatelecommunicationadministrationandarecognizedoperatingagency.

CompliancewiththisRecommendationisvoluntary.However,theRecommendationmaycontaincertainmandatoryprovisions(toensure,e.g.,interoperabilityorapplicability)andcompliancewiththeRecommendationisachievedwhenallofthesemandatoryprovisionsaremet.Thewords"shall"orsomeotherobligatorylanguagesuchas"must"andthenegativeequivalentsareusedtoexpressrequirements.TheuseofsuchwordsdoesnotsuggestthatcompliancewiththeRecommendationisrequiredofanyparty.

INTELLECTUALPROPERTYRIGHTS

ITUdrawsattentiontothepossibilitythatthepracticeorimplementationofthisRecommendationmayinvolvetheuseofaclaimedIntellectualPropertyRight.ITUtakesnopositionconcerningtheevidence,validityorapplicabilityofclaimedIntellectualPropertyRights,whetherassertedbyITUmembersorothersoutsideoftheRecommendationdevelopmentprocess.

AsofthedateofapprovalofthisRecommendation,ITUhadnotreceivednoticeofintellectualproperty,protectedbypatents/softwarecopyrights,whichmayberequiredtoimplementthisRecommendation.However,implementersarecautionedthatthismaynotrepresentthelatestinformationandarethereforestronglyurgedtoconsulttheappropriateITU-TdatabasesavailableviatheITU-Twebsiteat

/ITU-T/ipr/

.

?ITU2021

Allrightsreserved.Nopartofthispublicationmaybereproduced,byanymeanswhatsoever,withoutthepriorwrittenpermissionofITU.

Rec.ITU-TY.2501(09/2021)iii

TableofContents

Page

1Scope

1

2References

1

3Definitions

1

3.1Termsdefinedelsewhere

1

3.2TermsdefinedinthisRecommendation

1

4Abbreviationsandacronyms

1

5Conventions

2

6Backgroundandmotivations

2

6.1Fromtheperspectiveofcomputingpowerresourceconsumers

2

6.2Fromtheperspectiveofcomputingpowerresourceproviders

3

6.3Fromtheperspectiveofnetworktechnology

3

6.4Fromtheperspectiveofabusinessmodel

3

7Scenarios

4

7.1Scenariosoflowlatency

4

7.2Scenariosofhighmobility

5

8Requirementsofcomputingpowernetwork

5

9Generalframeworkofthecomputingpowernetwork

6

10Functionalarchitectureofcomputingpowernetwork

7

10.1CPNresourcelayer

7

10.2CPNcontrollayer

8

10.3CPNservicelayer

8

10.4CPNorchestrationandmanagementlayer

9

11Securityconsiderations

9

Rec.ITU-TY.2501(09/2021)1

RecommendationITU-TY.2501

Computingpowernetwork–Frameworkandarchitecture

1Scope

ThisRecommendationprovidestheframeworkandarchitectureofthecomputingpowernetwork,specifiesitsfunctionalentitiesanddefinesthefunctionalitiesofthesefunctionalentities,inadditionthisRecommendationalsoprovidesgeneralscenarios,requirementsandsecurityconsiderationsofthecomputingpowernetwork(CPN).

2References

ThefollowingITU-TRecommendationsandotherreferencescontainprovisionswhich,throughreferenceinthistext,constituteprovisionsofthisRecommendation.Atthetimeofpublication,theeditionsindicatedwerevalid.AllRecommendationsandotherreferencesaresubjecttorevision;usersofthisRecommendationarethereforeencouragedtoinvestigatethepossibilityofapplyingthemostrecenteditionoftheRecommendationsandotherreferenceslistedbelow.AlistofthecurrentlyvalidITU-TRecommendationsisregularlypublished.ThereferencetoadocumentwithinthisRecommendationdoesnotgiveit,asastand-alonedocument,thestatusofaRecommendation.

[ITU-TY.2011]RecommendationITU-TY.2011(2004),Generalprinciplesandgeneral

referencemodelforNextGenerationNetworks.

[ITU-TY.2012]RecommendationITU-TY.2012(2010),Functionalrequirementsand

architectureofnextgenerationnetworks.

3Definitions

3.1Termsdefinedelsewhere

ThisRecommendationusesthefollowingtermsdefinedelsewhere:

None.

3.2TermsdefinedinthisRecommendation

ThisRecommendationdefinesthefollowingterms:

None.

4Abbreviationsandacronyms

ThisRecommendationusesthefollowingabbreviationsandacronyms:

5G5th-Generation

AI

CPN

OAM

SDN

ArtificialIntelligence

ComputingPowerNetwork

OperationAdministrationandMaintenance

SoftwareDefinedNetwork

UPF

V2X

UserPlaneFunction

VehicletoX

vBRASvirtualBroadbandRemoteAccessServer

2Rec.ITU-TY.2501(09/2021)

vCPEvirtualCustomerPremiseEquipment

VRVirtualReality

5Conventions

InthisRecommendation:

Thekeywords"isrequiredto"indicatearequirementwhichmustbestrictlyfollowedandfromwhichnodeviationispermittedifconformancetothisRecommendationistobeclaimed.

Thekeywords"isrecommendedto"indicatearequirementwhichisrecommendedbutwhichisnotabsolutelyrequired.Thus,thisrequirementneednotbepresenttoclaimconformance.

6Backgroundandmotivations

Thecomputingpowernetwork(CPN)isanewtypeofnetworkthatrealizesoptimizedresourceallocation,bydistributingcomputing,storage,networkandotherresourceinformationofservicenodesthroughanetworkcontrolplane(suchasacentralizedcontroller,distributedroutingprotocol,etc.).CPNcombinesnetworkcontextanduserrequirementstoprovideoptimaldistribution,

association,transactionandschedulingofcomputing,storageandnetworkresources.CPNprovidestechnicalopportunitieswithinacertainbusinesscontext.

6.1Fromtheperspectiveofcomputingpowerresourceconsumers

Demandsforcomputingpowerareincreasingdaybydayandubiquitouscomputingpowerresourcesareneeded.

Themarketforcomputing-intensiveservices,suchastheAI-basedservicesisgrowing.Thelarge-scaleapplicationofartificialintelligence(AI)algorithmsacrossvariousindustriesrequiresagreatamountofcomputingpowerresources.Inthistheageofcloudcomputing,thetendencyistobuildcentralizedcomputingpowerresourcespoolstosolvethisproblem.However,insomenewscenarios,thesecentralizedpoolsdonotsatisfyrequirementsasappropriatecomputingpowernodesneedtobeselectedaccordingtotheservicecharacteristics,price,andnetworkconditions.Anaccesscontrolsystembasedonfacerecognition,showninFigure6-1,servestoexplainwhyubiquitouscomputingpowerresourcesareneeded.

Figure6-1–Accesscontrolsystembasedonfacerecognition

Rec.ITU-TY.2501(09/2021)3

Inthisexample,anAItrainingpartcanbedeployedincentralizedcomputingpowerpools,suchasacloudcomputingcentre,toperformcomplexcomputingprocessesoffline.However,inthereasoningphase,suchasforinstance,whenrecognizingafacetoopenadoor,thelatencyfactorshouldbeconsidered.Iftheimageorvideoinformationissenttoacloudcomputingcentreforprocessing,thewaitingtimemaybetoolongtobepractical.Consequently,servicesneedtoconsidernotonlycomputingpowerresourcesbutalsothetimeforprocessing,thetimefortransmittingandthecost.Forfully-centralizedcomputingpowerresourcepools,theycannotsatisfyallrequirements,whiledistributedcomputingpowerresourcepools,suchasedgecomputingnodes,haveobviousadvantagesinbandwidthandlatency,thusmeetingthedevelopmentrequirementsoffutureservices.

6.2Fromtheperspectiveofcomputingpowerresourceproviders

Theconstructioncostsofcomputingpowerresourcesperunitaredecreasing.

InaccordancewithMoore'sLaw,theperformanceofchipsisimproving,whilethecostsofcomputingpowerresourcesperunitaredecreasing.Comparedwiththeprovisionofelectricity,theprovisionofcomputingpowerisrelativelyeasy.Intheory,aslongasthereisspaceandelectricityandduecompliancewithfiresafetyregulations,allunitsorevenindividualscanbuildacomputingpowerresourcepooltoprovidecomputingresourcesforvariousservices.However,theexistingcomputingresourcespools,self-builtbyenterprises,aregeographicallyscattered,andtheirutilizationisquitelow.Ifalltheseresourcescouldbeused,itwouldgeneratemorevalueandcreatemorewealthforusersandsociety.Itisalsonotedthatsuper-computingcentresownalargeamountofcomputingpowerresources,whileusersdonothaveawaytoapplyfortheseresources.Consequently,thereisaneedtofindanewsolutionandtechnologysystemtorealizecomputingpowerresourcessharing.

6.3Fromtheperspectiveofnetworktechnology

Networkdevelopmenthaslaidthefoundationfortheflexibleprovisionofmulti-partycomputingresources.

Forsharingofcomputingpowerresources,theaccessibilityofcomputingpowerresourcesneedstobeconsideredfirst.Withthedevelopmentofnetworktechnologiessuchas5Gandsoftwaredefinednetworks(SDNs)thenetworkisnownolongeracauseofbottlenecks.Usersandcomputingpowerresourcescanbeefficientlyconnectedthroughnetworks.

Theisolatedcomputingpowerpoolisdifficulttotransformintoameasurableandsaleablecommoditythatcanpromoteasustainabledevelopmentoftheindustrychain.Withthedevelopmentofnewnetworktechnologiessuchas5G,computingpowerresourcescanbeconnectedinasimple,efficient,andlow-costway,andcanbesoldandusedthroughanewbusinessmodel.

6.4Fromtheperspectiveofabusinessmodel

Anewbusinessmodelisnecessary.

Howtointegratetheresourcesofallparties,sothatuserscanflexiblychooseappropriatecomputingresourcesaccordingtoservicerequirementsisproblematic.Cloudcomputingprovidersneedtoemployalotofpeopletonegotiatewithownersofcomputingpowerresources,signandthenexecuteservicecontractsinordertointegratethediscretecomputingpowerresources.Inaddition,operationandmaintenanceoperationsalsorequireagreatnumberofstaff.Thisgoesagainsttheirprinciplesofminimalistoperationandmaintenance.Consequently,cloudserviceprovidersprefertocooperatewithenterpriseswithlargeamountsofresourcesratherthansmallcompaniesorevenindividuals.

4Rec.ITU-TY.2501(09/2021)

Itishardfortheownersofsmallandmid-sizedcomputingresourcespoolstocompetewithlarge-scalecloudcomputingserviceproviders.Firstofall,duetothelimitationsoftheirtechnicalcapabilities,theproductsandservicesthatcanbeprovidedarequitelimited.Secondly,limitedbythesizeoftheirenterprises,theirmarketingeffortsarelimited,anditisthereforedifficulttoinformpotentialusersoftheirresourceinformation.

Withthesupportofnewtechnologies,computingpowerresourcesinformationcanbedistributedthroughnetworks.Acomputingpowertransactionplatformforresourceprovidersandresourceconsumerscouldbeprovided.Throughthiscomputingpowertransactionplatform,computingpowerprovidersandconsumersdonotneedtobeconcernedaboutthetransactionpartners.Thetransactionplatformshieldsthedifferencebetweentheprovidersandtheconsumers.Providersandconsumersdonotperceiveeachother,andtheycanpayaccordingtounifiedrules,forminganewresourcetransactionmodel.

7Scenarios

Computingpowernetworksmainlyworkintwotypesofscenarios:thefirsttypearelow-latencyscenariosrepresentedbyedgecomputing.Inthiskindofscenario,thecomputingpowernetworkneedstoprovidecoordinationbetweencomputingresources;thesecondtypearehigh-mobilityscenariosrepresentedbylivebroadcastandvehicletoX(V2X).Inthiskindofscenario,thecomputingpowernetworkneedstoprovideflexibleschedulingforservices.

7.1Scenariosoflowlatency

Thelowlatencyrequirementofedgecomputingneedstheinterworkingofcomputingandnetworkcapabilities.Lowlatencyisoneofthemostcriticalcharacteristicsofedgecomputing,anditisalsothekeyfactorthatdistinguishesedgecomputingfromcloudcomputing.Intheedgecomputingscenario,clientshaveanurgentneedforlowlatency,highcomputingcapabilitiesandnetworkingroutingcapabilities,whichisalreadybeyondthecapabilitiesoftraditionalcloudserviceproviders,andnetworkcapabilitieshavebecomeapartofthecloudcomputingcapabilities.Therefore,inadditiontodeployingcomputingnodesattheedgeofthenetwork,italsoneedsa'computing-networkinterworking'reconstructionofthesubstratenetworkarchitecture.Thisreconstructionevolvesthetraditionalnetworkarchitecturemainlycarryingnorth-southtraffictothenewnetworkarchitecturewhichcanscheduleflexiblytakingintoaccountlatencyandcarryeast-westtrafficaswell.

Thenewapplicationofedgecomputingneeds'cloud-networkinterworking'networkarchitecture.ForatypicalAIapplicationforexample,AIreasoningneedslowlatency,thereforeitshouldbedeployedinedgecomputingnodes.WhileAItrainingcanbedeployedinacentralizedcloudcomputingplatformatlowcostasitneedshighcomputingcapabilitiesbuthasnorequirementsonlatency.However,thecurrentedgecomputingsolutiongenerallyfollowsthenetworkconstructionschemeoftraditionaldatacentres.Thisnetworkconstructionsolutionconsistsoftwolayersofswitches(aggregation+accessorSpine-Leafarchitecture)andonelayerofegressrouters[ITU-TY.2011]and[ITU-TY.2012].Thiskindofnetworkingarchitectureincreasesthenumberofroutesanddevicesoftheedgecomputingnodes.Italsoincreasesthelatency,andthelatencyisuncontrollablewhichisquitedifferentfromtheexpectedindicators.

Therefore,forAIapplications,computingpowernetworktechnologyisneededtocollecttheresourcesandnetworkinformationabouteachresourcenode,then,accordingtotheservicerequirement,theAItrainingtaskwiththeheaviestcomputingneedscanbedeployedonthecentralcloud.Thecomputingtaskssuchasfeatureextraction,templatematching,andobjectrecognitionaredynamicallyoffloadedtoappropriateedgenodes.Theterminalonlyneedstoberesponsiblefortargettrackingandscreenrenderinganddisplaytoreducetheend-to-endlatency.

Rec.ITU-TY.2501(09/2021)5

7.2Scenariosofhighmobility

ServicessuchaslivebroadcastfortravelandV2Xoftenhavehighmobility.Inanexamplescenariooflivebroadcast,anchorXisconductingatravellivebroadcast.Duringthelivebroadcast,theanchorneedstogotocityBfromcityAandsharethesceneryalongthewaybetweencityAandcityB.Livebroadcastservicesneedtobeabletoprovidesceneryfeatures,real-timeinteractionwiththeaudience(suchasbarragecommunication,voiceandvideoconnections,livestreamingcommerce,etc.).Inthefuture,VRlivebroadcastwillrequirestrongerimagerendering,greatercomputingpowerandlowerlatency.Asthelocationoftheanchorcontinuestochange,usingthesamecomputingresourcesthechangewillincreaseaswillthephysicaldistancebetweentheserviceandthecomputingresources,leadingtoincreasedlatency.Therefore,thecomputingpowernetworkmayselectmatchingresourcesforusersinreal-timebasedonthelocationandresourceconditions,soastoprovidetheaccompanyingresourcesandimprovetheuserexperience.AsshowninFigure7-1,whenanchorXisincityA,resourcepoolN1providesservices,whenanchorXiisincityB,resourcepoolN2providesservices.

Figure7-1–Livebroadcastscenario

8Requirementsofcomputingpowernetwork

Computingpowernetworksarerequiredtohavetheabilitytoallocatethecomputingpowerresourcesaccordingtouserrequirementsandadjusttheallocationstrategywiththechangeofrequirementsinatimelymanner.Thus,thefollowingrequirementsmustbesatisfied:

–

–

–

–

–

CPNisrequiredtosupportcomputingpowermeasurementforthequantificationof

computingpowerresources.Inotherwords,CPNisrequiredtoprovideunifieddescriptionforcomputingpowerresources.

CPNisrequiredtohavethecapabilitytocollectresourcesinformation.

CPNisrequiredtomakeresourceallocationstrategiesforusersaccordingtouserrequirements.

CPNisrequiredtoestablishnetworkconnectionbetweenresourceusersandresourceprovidersaccordingtotheresourceallocationstrategies.

CPNisrequiredtosupportresourcetransactionbetweenresourceusersandresourceproviders.

6Rec.ITU-TY.2501(09/2021)

–CPNisrequiredtosupportresourcemonitoringtoupdateresourceinformationinatimelymanner.

NOTE–Theunifieddescriptionreferstoasetofmeasurementrules,accordingtowhichwecancomparecomputingresourcesundercertainconditions.Itdoesnotmeanweneedtogiveonlyoneunitforallcomputingpowerresources.

9Generalframeworkofthecomputingpowernetwork

Acomputingpowernetworkframeworkincludesacomputingpowernetworkconsumer,acomputingpowernetworkprovider,acomputingpowernetworktransactionplatform,acomputingpowernetworkcontrolplane,anetworkoperator,etc.AnAIempoweredplatformcanbelinkedaccordingtoservicesdevelopmentrequirements.ThecomputingpowernetworkframeworkisshowninFigure9-1.

–

–

–

Figure9-1–Computingpowernetworkframework

Computingpowernetworkconsumer:Unitsorindividualswhoconsumecomputingpowerresourcesandnetworkresources.Theymakevariousrequirementsintermsofcost,performanceandsecurityaccordingtotheirservices.

Computingpowernetworkprovider:Referstotheunitsorindividualsthatcanprovidecomputingpowerresources.Thecomputingresourcepoolcanbeasmall-scaleedgecomputingnode,alargeormedium-sizedcloudcomputingnode,ametropolitancomputingnodeorasuper-computingcentre,etc.Therefore,theprovidercanbeatelecommunicationoperator,alargecloudserviceprovider,asmallormedium-sizedenterprise,asupercomputingcentre,orevenanindividual.

Computingpowernetworktransactionplatform:Aplatformthatallowscomputingpowernetworkprovidersandcomputingpowernetworkconsumerstomaketransactions.Itcanbe

Rec.ITU-TY.2501(09/2021)7

apublictransaction,thatis,thecomputingpowernetworkconsumersknowexactlywhoprovidesthecomputingpowerresources.Itcanalsobeananonymoustransactionmeaningthatthecomputingpowernetworkconsumersdonotneedtoknowwhothecomputingpowernetworkprovidersare,andthecomputingpowernetworktransactionplatformisresponsiblefortransactionreliabilityandcomputingsecurity.Inaddition,onthistransactionplatform,notonlycomputingresourcetransactions,butnetworkresourcetransactionsneedtobecompletedatthesametimeaccordingtolocationandservicerequirements.

–Computingpowernetworkcontrolplane:Itcollectscomputingpowerinformation,networkinformation,etc.,andsendsthisinformationtothecomputingpowernetworktransactionplatformforcomputingpowernetworkconsumerstoselectappropriatecomputingpowerresourcesandnetworkresources,andthenprovidesthecomputingpowernetworkconsumerswiththeoptimalcomputingpowerresourceallocationandnetworkconnectionsolution.

–Networkoperator:Anoperatorthatprovidesconnectionserviceswhichconnectusersandcomputingpowerresources,andcanprovidedifferentlevelsofconnectionservicesaccordingtouserrequirements.

–AIempoweredplatform:Asanadditionalmoduleinthecomputingpowernetworkframework,itcanprovidecomputingpowerapplicationsforcomputingpowernetworkconsumersaswellasAI-basedauxiliaryoperationsforcomputingpowernetworkproviders.

10Functionalarchitectureofcomputingpowernetwork

ThefunctionalarchitectureofthecomputingpowernetworkisdepictedinFigure10-1.Itconsistsoffourlayerswhichprovidetheaforementionedfunctionalitiesbyinteractingwitheachother.

Figure10-1–Computingpowernetworkfunctionalarchitecture

10.1CPNresourcelayer

CPNresourcelayeriswheretheresourcesresidethatareprovidedbycomputingpowernetworkprovidersandnetworkoperators.Thisincludesresourcestypicallyusedinresourcenodes(cloud

8Rec.ITU-TY.2501(09/2021)

computingnode,edgecomputingnode,etc.)suchascomputingresources(servers,etc.),networkresources(switches,routers,etc.),storageresources(storagedevices),andalsothedeployedservicesthatrunontheservers.Inthislayer,theresourcesarevarious,heterogeneous,andbelongtomanyproviders.Usingunifiedidentificationcanrealizetheunifiedauthenticationandresourceschedulingofdifferentvendorsandheterogeneouscomputingpowerresources.

10.2CPNcontrollayer

TheCPNcontrollayer(realizedbytheCPNcontrolplane)collectstheinformationfromtheCPNresourcelayer,andsendsittoaservicelayerforfurtherpro