邊緣計算白皮書：云計算已死邊緣計算當(dāng)立-IEC

1、White PaperEdge intelligenceExecutive summaryThe cloud is dead long live the cloud!Driven by the internet of things (IoT), a new computing model edge-cloud computing is currently evolving, which involves extending data processing to the edge of a network in addition to computing in a cloud or a cent

2、ral data centre. Edge-cloud computing models operate both on premise and in public and private clouds, including via devices, base stations, edge servers, micro data centres and networks.Expansion of the IoT and digital transformation will generate entirely new businesses and markets, with both vend

3、ors and customers creating new demands on computing and networking infrastructures across all industries (automotive, aerospace, life safety, medical, entertainment and manufacturing, to name just a few).Edge intelligence (EI) is edge computing with machine learning (ML) and advanced networking capa

4、bilities. This means that several information technology (IT) and operational technology (OT) industries are moving closer towards the edge of the network so that aspects such as real- time networks, security capabilities to ensure cybersecurity, self-learning solutions and personalized/customized c

5、onnectivity can be addressed.Container technology and ML are the leading technological responses to the demands of these new businesses and markets. Bringing these technologies to the edge, will fulfil the promise of EI.Likewise, fifth generation wireless technology (5G) is bringing together IT and

6、telecommunications,e.g. by enabling data centres at the edge of networks as well as the possibility of implementingindustry-specific networks enabled by virtualization and software-defined networking principles in a single environment. Most of the anticipated 5G applications are driven by the IoT.Ba

7、sed on a thorough review of such developments, this White Paper offers the following conclusions:Containerization will be important for EI, but no specific Standards exist in this area, although many open source initiatives have been developed, such as Docker and the Open Container Initiative (OCI).

8、Common data models for edge computing node (ECN) communication are essential to the success of EI.Micro data centres will become more important, for a number of reasons, including the ability to provide low latency and to process large volumes of data, thereby avoiding their transportation to the cl

9、oud.5G networks will provide data centres at the edge and the possibility to implement industry- specific networks enabled by virtualization and software-defined networking principles.The best user interface is no user interface. As IoT makes manual data input largely obsolete and ML and artificial

10、intelligence (AI) take over decision-making, this becomes a reality.This White Paper formulates recommendations based on a review of use cases versus existing technology and Standards. Additionally, this White Paper suggests that all recommendations be demonstrated and supported by the use of a test

11、bed in collaboration with the Industrial Internet Consortium (IIC) to complement Standards with open source implementations.Executive summarySpecifically, the present White Paper makes the following recommendations to industry:Prepare for disruption on business and commercial models.Utilize 5G Stand

12、ards to facilitate edge computing and EI solutions.Include micro data centres in EI solution architecture, ideally employing containerization.Agree on a common approach to orchestration and lifecycle management to avoid market fragmentation.Agree on a common approach to ML (tools, model implementati

13、on) to avoid market fragmentation.It is also recommended that the IEC take a larger role in promoting the development of the software component of electrotechnical systems. The IEC is in a unique position to drive EI forward.Accordingly, this White Paper also outlines specific recommendations for th

14、e IEC Standardization management Board (SMB) in the areas of:Credibility and (decentralized) trustSelf-organization, self-configuration and self- discoveryImplementation of algorithms for MLThere already exist many standardization and consortium activities related to edge computing. This situation p

15、rovides challenges with regard to optimizing edge computing standardization, but also opportunities to create a more positive standardization ecosystem that will support the needs of governments, industry and users.Finally, such an ecosystem should be one of collaboration across the spectrum of stan

16、dards development organizations (SDOs) and consortia. Specifically, this White Paper recommends that the IEC collaborate with IIC to complement Standards with open source implementations and testbeds. These include covering horizontal, vertical and specialty Standards, as well as EI Standards.So how

17、 will the current technology ecosystem be influenced by these new concepts and the emerging technology? This White Paper provides insights and recommendations on various aspects of EI.Section 1 gives a brief description of this new technology by highlighting the benefits and opportunities, as well a

18、s the challenges and foreseen Standards, enabling the development of the true potential of EI. The section concludes with a rough classification of edge computing scenarios.Section 2 elaborates further on the evolution of computing models, how the IoT disrupts todays cloud-based network architecture

19、, and how that disruption calls for a new computational model “on top of the cloud”. Moreover, Section 2.4 provides a high-level novel edge computing architecture for EI.Having established a general understanding of what EI is, section 3 provides a deeper insight on the trend drivers and needs match

20、ing EI in different industries. Moreover, an overview of what constitutes state of the art regarding the evolution of hardware, software and architectures for communication networks is included in this section to capture technological synergies, e.g. virtualization and the 5G system, and emerging te

21、chnologies such as ML and blockchains.Section 4 briefly covers a set of use cases from the manufacturing, smart cities, smart building and life safety areas, highlighting their requirements, observed gaps, needed capabilities for overcoming such gaps and the standardization topics that can be pursue

22、d in this regard. A detailed description of the use cases can be found in Annex A.In section 5, further details are provided concerning specific gaps in the various use cases. The section concludes with a framework for discussion focusing briefly on common gaps between the analyzed EI use cases.Exec

23、utive summaryIn section 6 the needed capabilities for all the use cases are extracted with a more detailed presentation of how the use cases would benefit from employing these capabilities.In section 7 the identified missing Standards are described and an analysis on common missing Standards regardi

24、ng the use cases is given.In section 8, as a result of the use case analysis, a fusion of some of the use cases are described by the means of a testbed, along with the possible extensions.Section 9 concludes this White Paper by presenting recommendations to industry concerning the possible benefits

25、to be gained from employing EI, a series of general recommendations to SDOs and, finally, recommendations targeted specifically at IEC members.AcknowledgmentsThis White Paper has been prepared by the edge intelligence project team, in the IEC Market Strategy Board (MSB), with major contributions fro

26、m the project partner, the Fraunhofer Institute for Open Communication Systems FOKUS, and the project leader, Huawei Technologies Co. Ltd. The project team met four times November 2016 (Berlin, Germany), January 2017 (Macau), March 2017 (Berlin, Germany) and May 2017 (Palo Alto, US). The project tea

27、m is directed by Mr. Xuemin Wang, Director of the Industry Standards Department, Huawei Technologies Co. Ltd, a MSB member. The project team is listed below:Mrs. Andreea Ancuta Corici, Fraunhofer FOKUS, Project Partner LeadMr. Yun Chao Hu, Huawei, Project Manager Prof. Dr. Thomas Magedanz, Fraunhofe

28、r FOKUS Mr. Marc Emmelmann, Fraunhofer FOKUSDr. Andreas Vogel, SAPMr. Tetsushi Matsuda, Mitsubishi Electric Mr. Michael Cronin, Johnson Controls Inc. Mr. Victor Kueh, HuaweiMr. Bin (Remy) Liu, Huawei Mr. Yaling Zhou, Huawei Mr. Meigang Liu, Huawei Mr. Yang Shi, HuaweiMr. Kunlun Lu, HuaweiDr. Marius

29、Corici, Fraunhofer FOKUS Dr. Mikhail Smirnov, Fraunhofer FOKUSDr. Alexander Willner, Fraunhofer FOKUSMr. Christopher Stephan Kertess, Fraunhofer FOKUSMr. Peter Lanctot, IEC Dr. Gilles Thonet, IEC 5Table of contentsList of abbreviations10Glossary16Section 1Introduction19Scope of this White Paper20Ben

30、efits and opportunities21Common challenges of edge computing scenarios21Section 2The evolution of computing models towards edge computing25Shared and central resources versus exclusive and local computation25IoT disrupts the cloud25Characteristics of the new computing model26Blueprint of edge comput

31、ing intelligence26Definition and high level architecture26Application areas27Section 3Trend drivers and state of the art for edge intelligence33Industry needs33Buildings and life safety industry35Smart manufacturing38Automotive38Information and communication technology41Security domains42Blockchain4

32、3Hardware evolution50Data centre evolution50IoT gateways/edge servers50Smart sensors/end nodes51Software evolution53IoT edge computing53ECN OS54Containerization and microservices56Machine learning59Architecture613.4.1Data collection architectures613.4.2oneM2M623.4.3IIC architecture673.4.4OPC-UA arch

33、itecture683.4.5Core networks683.4.6Access networks773.4.7Programmable infrastructures79Section 4Use cases and requirements for edge intelligence83Use cases overview83 HYPERLINK l _TOC_250060 Framework for discussion85Section 5Technology gaps89 HYPERLINK l _TOC_250059 Factory productivity improvement

34、89 HYPERLINK l _TOC_250058 Connected city lighting89 HYPERLINK l _TOC_250057 Smart elevator90 HYPERLINK l _TOC_250056 Indoor location tracking91 HYPERLINK l _TOC_250055 Lone worker safety91 HYPERLINK l _TOC_250054 Access control tailgating detection/fire detection via surveillance cameras91Section 6

35、Needed capabilities93 HYPERLINK l _TOC_250053 Integration of edge and core93 HYPERLINK l _TOC_250052 Edge services93 HYPERLINK l _TOC_250051 Factory improvement productivity93 HYPERLINK l _TOC_250050 Connected city lighting94 HYPERLINK l _TOC_250049 Smart elevator95 HYPERLINK l _TOC_250048 Indoor lo

36、cation tracking95 HYPERLINK l _TOC_250047 Lone worker safety96 HYPERLINK l _TOC_250046 Access control tailgating detection/fire detection via surveillance cameras96Section 7Standards and role of open source97 HYPERLINK l _TOC_250045 Standards for self-organization, self-configuration, self-discovery

37、97 HYPERLINK l _TOC_250044 Trust/ decentralized trust98 HYPERLINK l _TOC_250043 Credible information98 HYPERLINK l _TOC_250042 E/W communication Standards between multiple ECNs99 HYPERLINK l _TOC_250041 Containerization Standard for embedded systems99Open standard for implementation of algorithm for

38、 machine learning100 HYPERLINK l _TOC_250040 Comprehensive standard tackling carrier mode selection in case of loss of connectivity101 HYPERLINK l _TOC_250039 Role of open source101Section 8Edge intelligence use case testbed103 HYPERLINK l _TOC_250038 Potential use cases which can be implemented on

39、the testbed103 HYPERLINK l _TOC_250037 Technology building blocks104 HYPERLINK l _TOC_250036 Future developments104Section 9Conclusions and recommendations105 HYPERLINK l _TOC_250035 Conclusions105 HYPERLINK l _TOC_250034 Recommendations106 HYPERLINK l _TOC_250033 Industry-targeted recommendations10

40、6 HYPERLINK l _TOC_250032 General recommendations106 HYPERLINK l _TOC_250031 Recommendations addressed to the IEC and its committees107Annex AUse cases and requirements for edge intelligence109 HYPERLINK l _TOC_250030 Factory productivity improvement109 HYPERLINK l _TOC_250029 Scope109 HYPERLINK l _

41、TOC_250028 Objectives109 HYPERLINK l _TOC_250027 Narrative of use case109Use case conditions109Further information for the use case109 HYPERLINK l _TOC_250026 Connected city lighting111 HYPERLINK l _TOC_250025 Scope111 HYPERLINK l _TOC_250024 Objectives111 HYPERLINK l _TOC_250023 Narrative of use ca

42、se111Use case conditions113 HYPERLINK l _TOC_250022 Further information for the use case113 HYPERLINK l _TOC_250021 Smart elevator116 HYPERLINK l _TOC_250020 Scope116 HYPERLINK l _TOC_250019 Objectives116Narrative of the use case116Use case conditions117 HYPERLINK l _TOC_250018 Further information f

43、or the use case118 HYPERLINK l _TOC_250017 Indoor location tracking119 HYPERLINK l _TOC_250016 Scope119 HYPERLINK l _TOC_250015 Objectives119 HYPERLINK l _TOC_250014 Narrative of the use case119 HYPERLINK l _TOC_250013 Use case conditions120 HYPERLINK l _TOC_250012 Lone worker safety121 HYPERLINK l

44、_TOC_250011 Scope121 HYPERLINK l _TOC_250010 Objectives121 HYPERLINK l _TOC_250009 Narrative of the use case121Use case conditions122 HYPERLINK l _TOC_250008 Access control tailgating detection124 HYPERLINK l _TOC_250007 Scope124 HYPERLINK l _TOC_250006 Objectives124 HYPERLINK l _TOC_250005 Narrativ

45、e of the use case124Use case conditions124 HYPERLINK l _TOC_250004 Fire detection via surveillance cameras126 HYPERLINK l _TOC_250003 Scope126 HYPERLINK l _TOC_250002 Objectives126 HYPERLINK l _TOC_250001 Narrative of the use case126Use case conditions126 HYPERLINK l _TOC_250000 Bibliography129List

46、of abbreviationsAAAauthentication, authorization and accountingADautomated drivingADNapplication dedicated nodeAEapplication entityAFapplication functionAIartificial intelligenceANDSFaccess network discovery and selection functionAPIapplication programming interfaceASapplication serverASNapplication

47、 service nodeBLEBluetooth low energyBSSbusiness support systemCapExcapital expenditureCCIco-channel interferenceCDNcontent delivery networkC-ITScooperative intelligent transportation systemCNCcomputer numerical controlCNNconvolutional neural networkCPcontrol planeCPUcentral processing unitCRUDNcreat

48、e, retrieve, update, delete, notifyCSEcommon services entityCSFcommon services functionDCSdistributed control systemDNNdeep neural networkDSRCdedicated short-range communicationsDTLSdatagram transport layer securityD2Ddevice-to-device communicationTechnical and scientific termsECNedge computing node

49、EIedge intelligenceEMSelement management systemeNodeBevolved node BEPCevolved packet coreERPenterprise resource planningE/Weast/westFCAPSfault, configuration, accounting, performance and securityFDDfault detection and diagnosticsGISgeographic information systemGPSglobal positioning systemGPUgraphics

50、 processing unitHMMhidden Markov modelHPSLhigh pressure sodium lampHSShome subscriber serverIACSindustrial automation and control systemIAMidentity and access control managementIIoTindustrial internet of thingsINinfrastructure nodeIoTinternet of thingsIPEinterworking proxy entityIRCintelligent and r

51、esilient controlISGindustry specification group (ETSI)ISMindustrial, scientific and medicalISPinternet service providerITinformation technologyJARJava archiveLEDlight-emitting diodeL1/L2level-1/level-2 (cache)LTElong-term evolutionLWM2Mlightweight machine-to-machine protocolMANOmanagement and orches

52、trationMECmulti-access edge computingMESmanufacturing execution systemsMImachine intelligenceMLmachine learningMMEmobility management entityMNmiddle nodeM2Mmachine-to-machine NB-IoTnarrowband IoT NASnon-access stratumNCnumeric controlNFnetwork functionNFVnetwork function virtualizationNIDDnon-IP dat

53、a deliveryNoDNnon-oneM2M device nodeN/Snorth/southNSDnetwork service descriptorNSEnetwork services entityODPopen distributed processingO&Moperations and maintenanceOAMoperations, administration and maintenanceOpExoperational expenditureOSoperating systemOSSoperations support systemOToperational tech

54、nologyPCpersonal computerPCRFpolicy and charging rules functionPDN-GWpacket data network (pdn) gateway (gw)PKIpublic key infrastructurePLCprogrammable logic controllerPNFphysical network functionPoPpoint of presenceProSeproximity servicesPUFphysical unclonable functionQoSquality of serviceRAMrandom

55、access memoryrktRocket container technology RNISradio network information services RSUroadside unitSCEFservice capability exposure functionSDNsoftware-defined networkSDSsoftware-defined storageSGWserving gatewaySIMsubscriber identity moduleSMSshort message serviceSoCsystem on a chipSONself-organizin

56、g network telecomtelecommunication TLStransport layer securityTOFtraffic offload functionTSNtime-sensitive networkingUEuser equipmentUIuser interfaceUICCuniversal integrated circuit cardUNBultra narrow bandUPuser planeUWBultra wide bandVLANvirtual local area network VNFvirtual network function VoIPv

57、oice over internet protocol VPNvirtual private networkVRUvulnerable road userV2Ivehicle-to-infrastructureV2Xvehicle-to-everythingWAVEwireless access in vehicular environmentsWLANwireless local area networkXMLextensible markup languageOrganizations, institutions,companies and protocols3GPP3rd Generat

58、ion Partnership Project5G PPP5G Infrastructure Public Private PartnershipAMQPadvanced message queuing protocolANIMA Autonomic Networking Integrated Model and Approach (IETF)ARIBAssociation of Radio Industries and Businesses (Japan) ATISAlliance for Telecommunications Industry Solutions (US) CCSA Chi

59、na Communications Standards AssociationCoAP constrained application protocolcuDNN CUDA deep neural networks library (NVIDIA)ETSIEuropean Telecommunications Standards InstituteHTTP hypertext transfer protocolIDCInternational Data CorporationIEEEInstitute of Electrical and Electronics Engineers IEETIn

60、stitute of Ethics and Emerging Technologies IETFInternet Engineering Task ForceIICIndustrial Internet ConsortiumIPinternet protocolIpv4internet protocol version 4ISOInternational Organization for StandardizationLPWANIPv6 over Low Power Wide-Area Networks (IETF working group)MQTT message queue teleme