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)，請進(jìn)行舉報(bào)或認(rèn)領(lǐng)
文檔簡介
ConsiderationofAIbasedTelecomNetwork
ZhangHao
ChinaMobile
2024.3
2
ChinaMobileistheworld'sleadingcommunicationsserviceprovider
Networkscaleranks
1stintheworld
2023
1.9+
million
2023
370
million
Householdswithgigabitcoverage
5Gbasestations
Customerscaleranks
1stintheworld
2023
3.19
billion
1.69
Connections
(things)billion
Total
connections
2023
Incomescaleranks
1stintheworld
2023.10
Operatingrevenue
139,597
million
2023
Profit
14.718
billion
Threerevolutionstagesofcommunicationnetworks
NetworkIP-lisedrevolution
NetworkIT-lised
NetworkAI-lisedrevolution
revolution
2000s2010s2020s2030s
Narrowbanddigitalvoice
low-speeddata
High-speeddatainbroadbanddigitalvoice
Peopletopeople--Peopletothings--ThingstothingsCommunicationsInternetserviceprovider
Intelligence,universalconnectivity,elemental
Businessdemand
Technicalfeatures
Typicaltechnology
ExplosivegrowthinInternetbandwidth
andsubscribernumbers.Transformation
fromsinglevoiceservicetoconverged
service.
Fromcircuitswitchingtopacket
switching
FromswitchingIPtoend-to-endfullIP
IP、MPLS、SRv6
PTN
Softswitch、IMS、VoLTE
Growingdemandforbusiness
diversityandresilience
ITrapidlychangingtheshape
ofthenetwork.
FromClosedNetworkstoOpen
Business
FromRigiditytoResilience
SDN
NFV
SBA
convergence
AINeedsNetworksforEfficient
CommunicationPerformance.
NetworksneedAItoenhancehigher-order
self-intelligencecapabilities
NetworkforAI
AIforNetwork、AINative
RoCE、NVLink、GSE
MLOps
FederalLearning
3
Communicationnetworkispromptingthenewinformationtechnologytoallareasofdeeppenetration
4
Theeraofuniversalintelligencerequires
efficientcommunicationperformanceof
networks.
Majorshiftfrommobilecommunicationstomobileinformationservices
Intelligent
computing
infrastructure’s
enhancement
NetworkforAI
StrengtheningComputibilitywithNetworks
DistributedtrainingofGPUclustersbringsalarge
amountofcommunicationoverhead,andnetwork
performancebecomesabottleneckrestrictingAI
arithmeticenhancement
WideAreaServiceUniversality
Intelligence
Inclusive
Ubiquitous
Network
intelligence
generatednatively
AINative
AIgeneratednatively
The6Gmobileinformationnetworkwillprovide
thewholeprocessofinformationflowservices,
achievingthebasicplatformofAIubiquityand
universality.
Majorrequirementsandchallengesofnetworkintelligence
TheconvergenceofnetworkandAIincludestwoaspects:"networkenablesAI"and"AIempowers
network".ThefirstistoprovideanuniversalaccesstoAIservices.ThesecondisAI-enablednetworks
toimprovenetworkoperationandO&Mefficiency.
Networksneedtobequicklyadaptedto
thecustomisedrequirementsofdiverse
scenarios
Networkcomplexity
increasesfrom
generationto
generation
AIfornetwork
Operationsandmaintenance
AIisthekeypathtomeetthenewmetricsof
mobilecommunicationnetworks,empowering
networkstoimprovenetworkoperation
efficiency.
Morecomprehensive
networkperformance
measures
5
AIforNetwork:Twomajorscenarios
Byextractingtheregularfeaturesofdataincomplexscenarios,
AIcanhelpthetelecomnetworkimproveefficiencyinmaintenanceandoperationscenarios
NetworkMaintenance
Aroundtheentirelifecycleofnetworkplanning,
construction,maintenance,optimization,and
operation,AIoptimizestheprocesstoachievethe
costreductionandefficiencyenhancement
AIreplacinghuman
Offline
analyzing
Lowercost
Richscenarios
Centralized
data
Maintenance+AI
Unifieddatainterface
Lowerpowerconsumption
NetworkOperation
AIreconstructsthenetworkoperationprocessto
achievethebestmatchingofnetworkresources,operatingefficiencyanduserexperience
Simplescenario
Discretedata
Poordata
standardization
Lowerpowerconsumption
AIreplacingequipment
Onlineprocessing
Operation+AI
Lowercost
6
ProcedureIntegration
AIforNetwork:Intelligentnetworkmaintenance
AsglobaloperatorscontinuetoevolvetowardsL4orhigher-levelAutonomousNetworks(ANs),
networkmaintenancemodeupgradesfromautomationtointelligence.AnintelligentsystemwithAI
modelsisbecomingthetrendoffuturetelecomnetworkmaintenance.
Theintelligentnetworkmaintenancesystem
Visualization
?VisualizationofFCAPSdata
?Visualizationofnetworktopology
IntelligentGuidance
?SmartQ&A
?Onboardingguidance
RoutineMaintenance
?In-depthinspection
?Healthassessment
FundamentalNetworkModels&Tools
SpecializedModels(Alarm/Log/Perf/…)
AutomationTools(SI/Config/Visualize)
ICTLargeLanguageModel
ModelOptimizationforICTScenario
ICTPrivateDomainKnowledgeData
LargeLanguageModel(LLM)Base
FaultDiagnosis
?Hiddendangerprediction
?Intelligentrootcaseanalysis
KeepLive&Recovery
?Disasterrecoveryassessment
?Automaticexecution
NetworkOptimization
?QualityImprovement
?Energyefficiencyoptimization
Visualizing
RegionalLevel
ApplicationLevel
UELevel
......
7
NSSF
ADRF
NEF
PCF
UDR
AF
Nudr
Nnef
NnwdafNaf
Npcf
Nssf
Nadrf
AMF
OAM
SMF
MFAF
UPF
RAN
Observing
Applicationtype
Serviceexperience
data
RANloadstatus
Controling
QOSAssurance
SneakerMarketing
......
AIforNetwork:Intelligentnetworkoperation
NWDAF(NetworkDataAnalyticsFunction)isintroducedtorealizeoptimalmatchofnetworkresource,
achievingthehighestnetworkefficiencyandbetteruserexperience.
Architecture
TypicalCase
NWDAF
NoamNamfNsmfNupfNdccfNmfaf
DCCf
N4
UE
NWDAFbasednetworkoperationintelligence
?ServiceRegistration:Servicearea,AnalyticsID.
?Datacollection:5GCNF、AF、OAM,real-timecustomizedcollection.
?AI/MLtraining:Providemachinelearningmodels.
?Inferenceperforming：Prediction/statistics/recommendation.
?AnalyticsFeedback:consumerdecision/action.
Initiation
Recognition
Perception&Identification
AI
Inspection
Action
Decision
Calculating
KPIstatistics
Sneaker
QoEexperience
......
8
AIforNetwork:Fourmainissues
Issue1:DifficultyinSceneIdentification
EnablingScenarios
Support
Drive
Issue3:Thedispersionofmodel
Issue2:Poordataquality
___________Data
Network
Availability
AIModel________________
Construct
Issue4:Weaknetworkusability
Computility
Cloud
+production-orientedAI
Corenetwork
+production-orientedAI
Wirelessnetwork
+production-orientedAI
Terminal
+production-orientedAI
9
NetworkforAI：Networkstrengthenscomputility,supporting
Management
Applicationservice
PaaS
Software
platform
Hardware
resource
SSD
Convergedstorage
Block/File/ObjectStorage
High-performancestorage
HDD
RoCEv2
GSE
IB
Losslessnetwork
?Diversecomputility：ProvideAsPU、NPUAIintelligentcomputility
?High-performancestorage:FacilitatesAImodeltraining
?Losslessnetwork:ImprovetheefficiencyofAIclusters
?Softwareplatform:Unifiedmanagement,schedulingandabstractionforheterogeneousresources
AIdevelopment
BuildAIinfrastructure,provideAIcomputilityexposurecapabilitiesandempowerAIapplications.
Computilitynetworkmanagement,orchestrationandscheduling
Managementintelligence
O&Mmanagement
CoreNetwork
EPC/5GC/IMSsystem
Real-timetranslation
Digitalhumans
AIApplication
Fraud
prevention
Smart
recommendation
…
Modelasaservice
Basicmodel
Network
intelligence
G-PaaSA-PaaS
DB
LB
ServiceMesh
…
CI/CDpineline
Image
processing
Videoanalytics
Encryptdecrypt
Codecs
…
Unifiedmanagement,schedulingandabstractionplatformforhyperscaleheterogeneousresources
Ultra-lightweightvirtuz容aion
VMBMcontainer
Computilitynative
Cross-architecturecomputilityintegration
GPUpoolingMemorypooling
Computilitypooling
Diversecomputility
GPU
X86_64/ARM/RISC-V
DPUSmartNIC
DPU
NPUDSA
GP
U
CPU
①Infrastructurefor
②NetworkcapabilitiesforAI
?UbiquitousAIservices：AIvalue-addedservicesareprovidedbasedontheubiquity、mobilityfeaturesoftelecommunicationnetworks
?Modelservice:GeneratenetworkmodelsbasedonnetworkdataandprovideMaaSservice
ServiceEnablingLayer
ExternalAIService
Synaesthesia
Service
Taskdecompositioncapability+serviceorchestrationcapability
ComputingService
InternalAIService
ServiceFunctionLayer
Data
Plane
ComputingPlane
SafetyPlane
Control
Plane
UserPlane
Data
Management
Management
orchestration
Body
Capacity
openning
management
Autonomous
DigitalTwinBody
Scenemodellibrary
Twinlargemodel
Closed-loopprevalidatin
Unified
data
control
interface
Communicationandcomputinglayer
(Wirelesscommunication,opticalcommunication,computing,storage)
Intelligent
operationandmaintenancemanagement
resource
scheduling
Connectionandroutinglayer(Multi-access,trustedconnection,heterogeneous
inter
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，僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理，對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自己和他人造成任何形式的傷害或損失。
最新文檔
- 2025版智能垃圾分類處理項(xiàng)目深化履約服務(wù)合同3篇
- 2025年度股權(quán)及債權(quán)轉(zhuǎn)讓與清算事務(wù)處理合同
- 2025年度辦公房屋租賃合同二零二五年度租賃期限靈活
- 2025年度農(nóng)業(yè)科技項(xiàng)目投資合作合同范本
- 2025年度電力企業(yè)員工培訓(xùn)服務(wù)合同樣本
- 2025年度環(huán)保監(jiān)測服務(wù)合同變更管理范本
- 2025年度公司間高標(biāo)準(zhǔn)商務(wù)辦公樓租賃合同范本
- 二零二四年度藥品進(jìn)口代理集中招標(biāo)投標(biāo)合同3篇
- 2025年度窗簾市場渠道拓展合作合同4篇
- 2025年度回遷房項(xiàng)目配套設(shè)施建設(shè)監(jiān)管合同
- 2025福建新華發(fā)行（集團(tuán)）限責(zé)任公司校園招聘30人高頻重點(diǎn)提升（共500題）附帶答案詳解
- 山東鐵投集團(tuán)招聘筆試沖刺題2025
- 2025年中考英語總復(fù)習(xí)：閱讀理解練習(xí)題30篇（含答案解析）
- 陜西省英語中考試卷與參考答案(2024年)
- 北京市通州區(qū)市級名校2025屆高一數(shù)學(xué)第一學(xué)期期末考試試題含解析
- 小學(xué)生心理健康教育學(xué)情分析
- 超級大腦：孩子六維能力培養(yǎng)指南
- 2024年濰坊護(hù)理職業(yè)學(xué)院高職單招（英語／數(shù)學(xué)／語文）筆試歷年參考題庫含答案解析
- 顱腦損傷的生物標(biāo)志物
- 物流營銷（第四版） 課件 第一章 物流營銷概述
- 5A+Chapter+2+Turning+over+a+new+leaf 英語精講課件
評論
0/150
提交評論