边缘计算白皮书：云计算已死，边缘计算当立-IEC-2019.3-133页_3mb

Edge Intelligence White Paper Summary

Core Content

This White Paper explores the concept of Edge Intelligence (EI), which integrates machine learning (ML) and advanced networking capabilities into edge computing. It outlines the technological, market, and standardization developments that are shaping the future of computing models, especially in response to the Internet of Things (IoT) and digital transformation. The paper emphasizes the need for a new computing paradigm that operates at the edge of the network, complementing traditional cloud computing models.

Main Points

1. Evolution of Computing Models

• The traditional model of computing has oscillated between centralized resources and local computation.
• The IoT is driving a shift towards edge computing, where data processing occurs closer to the source, reducing latency and enabling real-time analytics.
• Edge-cloud computing combines both edge and cloud processing, with the former gaining increasing importance due to the need for low latency and local data handling.

2. Definition of Edge Intelligence

• Edge Intelligence (EI) is a model of edge computing enhanced by ML and advanced networking.
• It enables real-time networks, security capabilities, self-learning solutions, and personalized connectivity.
• EI is supported by 5G networks, which provide edge data centers and virtualized, software-defined networking environments.

3. Technology Drivers

• Containerization (e.g., Docker, OCI) is crucial for EI but lacks standardized specifications.
• Common data models for edge computing nodes (ECNs) are essential for interoperability and scalability.
• Micro data centres are becoming more important due to their ability to process large volumes of data locally, avoiding cloud transportation and latency issues.
• Machine learning is a key enabler for self-learning and decision-making at the edge.
• 5G enables edge computing and EI through low latency, high bandwidth, and network virtualization.

4. Use Cases and Requirements

• The paper identifies several use cases in manufacturing, smart cities, smart buildings, and life safety.
• These use cases require real-time processing, security, low latency, local decision-making, and context-awareness.
• Specific requirements include data aggregation, local analytics, network optimization, and standardization of communication protocols.

5. Technology Gaps

• The paper highlights gaps in factory productivity, connected city lighting, smart elevators, indoor location tracking, lone worker safety, and access control.
• These gaps relate to latency, security, interoperability, and standardization.
• A framework for discussing these gaps is provided, focusing on common challenges across use cases.

6. Needed Capabilities

• Integration of edge and core systems is necessary for seamless data flow and decision-making.
• ML algorithms, containerization, edge services, and network virtualization are key capabilities to be developed.
• These capabilities should be standardized and supported by open source and testbeds to ensure market consistency and innovation.

7. Standards and Open Source

• Standards are critical for interoperability, security, and scalability in EI.
• Common data models, orchestration, lifecycle management, and ML implementation standards are needed.
• Open source initiatives (e.g., Docker, OCI) are important for implementation, but standards are still lacking.
• The paper recommends collaboration between IEC and IIC to develop testbeds and open source implementations that complement existing standards.

8. Testbed Recommendations

• A testbed is proposed to demonstrate EI use cases and capabilities.
• The testbed should support horizontal, vertical, and specialty standards, as well as EI-specific standards.
• It should be used to validate and enhance the standardization process.

9. Conclusions and Recommendations

• The cloud is evolving, and edge intelligence is emerging as a critical component.
• The paper concludes with industry recommendations:
  • Prepare for disruption in business and commercial models.
  • Utilize 5G standards to enable edge computing and EI.
  • Include micro data centres in EI architectures.
  • Agree on common orchestration and lifecycle management standards.
  • Agree on common ML tools and models to avoid market fragmentation.
• IEC is encouraged to take a more active role in promoting EI standardization, especially in the software component of electrotechnical systems.
• A collaborative ecosystem across SDOs and consortia is recommended to support governments, industry, and users.

Key Information

• Edge intelligence combines edge computing with machine learning and advanced networking.
• 5G plays a vital role in enabling EI by providing edge data centers and virtualized networks.
• Containerization and open source are important for EI development but lack standards.
• Common data models and standardization are essential for interoperability and market growth.
• Micro data centres are becoming more important due to their low latency and local data processing capabilities.
• The Industrial Internet Consortium (IIC) and IEC are recommended to collaborate on testbeds and standardization efforts.

Recommendations

• Industry should:
  • Prepare for disruption in business models.
  • Use 5G standards to support edge computing and EI.
  • Integrate micro data centres into EI architectures.
  • Agree on common orchestration and ML standards to prevent market fragmentation.
• IEC should:
  • Take a larger role in promoting EI standardization.
  • Collaborate with IIC to develop testbeds and open source implementations.
• SDOs and consortia should:
  • Work together to create a collaborative standardization ecosystem.
  • Focus on horizontal, vertical, and specialty standards, including EI.

Structure of the White Paper

• Executive Summary: Provides an overview of EI and its importance.
• Acknowledgments: Credits the edge intelligence project team, Fraunhofer FOKUS, and Huawei.
• Table of Contents: Lists the sections and appendices.
• List of Abbreviations: Includes technical terms and acronyms.
• Glossary: Defines key terms such as AI, ML, 5G, ECN, and EI.
• Sections:
  • Section 1: Introduction and scope of EI.
  • Section 2: Evolution of computing models towards edge computing.
  • Section 3: Trend drivers and state-of-the-art for EI.
  • Section 4: Use cases and their requirements.
  • Section 5: Technology gaps in various use cases.
  • Section 6: Needed capabilities for EI.
  • Section 7: Standards and the role of open source.
  • Section 8: EI use case testbed and its potential.
  • Section 9: Conclusions and recommendations.
• Annex A: Detailed descriptions of the use cases.

Conclusion

Edge Intelligence represents a paradigm shift in how data is processed, stored, and analyzed. It is driven by IoT, 5G, and machine learning, and requires standardization, open source, and collaboration across industry, consortia, and standards bodies like IEC and IIC. The paper outlines a vision for EI and provides recommendations to ensure its successful implementation and market adoption.