Markdown Version | Session Recording

Session Date/Time: 08 Nov 2021 12:00

nmrg Session Minutes

Summary

The nmrg session covered updates on several drafts related to Intent-Based Networking (IBN) and Digital Twins, alongside the introduction of a new draft on an AI Framework for Network Management. Discussions highlighted the progress of existing IBN drafts towards IRTF chair review and the need for further refinements. Digital Twin presentations showcased an industrial IoT implementation and proposed new drafts on data collection methods and a reference architecture. Authors of several drafts requested research group adoption, prompting chairs to outline the process for soliciting community feedback on document maturity and relevance to the group's research agenda.

Key Discussion Points

• Administrative & NMRG Document Status

  • Standard IETF Note Well, privacy, and code of conduct reminders were given.
  • The role of IRTF (research) versus IETF (standardization) was reiterated.
  • draft-irtf-nmrg-ibn-concepts-definitions: Revision is under preparation to address IRTF chair review comments received in September. Authors plan offline discussions with the shepherd and IRTF chair, Collin St. Clair.
  • draft-irtf-nmrg-ibn-classification: An updated version has been proposed with minor comments pending. Authors are making final changes and expect a new version this week, followed by an offline discussion with the IRTF chair.
  • Future NMRG plans include resuming monthly virtual meetings, potentially organizing a workshop on "actual network and editing associates license," and consolidating existing research documents.

• Network Management Intent (NMI) Draft (draft-dan-nmrg-network-measurement-intent)

  • Presenter: Dan (China Mobile, presented by a colleague)
  • Updates: The primary update involves a revised procedure diagram that integrates the NMI policy model. This model now connects the translation, compliance assessment, and orchestration modules, enabling outputs from translation and assessment to be converted into policies/actions for orchestration.
  • Request: The authors requested research group adoption for the draft or to explore writing a draft on IBN use cases.
  • Discussion:
    • Chairs clarified the NMRG's "adoption" procedure, which involves assessing document maturity based on scientific completeness, technical soundness, document shape, and author/community support, followed by a mailing list discussion.
    • The draft was identified as a valuable IBN use case, sparking discussion on whether to pursue individual use case drafts or a consolidated document.
    • References: Charles suggested a MEF draft on SD-WAN application performance measurement. Robert Wilson recommended the network telemetry framework from the OPS area working group.
    • Suggestion: Laurent proposed explicitly connecting this use case to the IBN classification draft for better integration within the group's work.

• Interconnection Intents Draft (draft-lucente-nmrg-interconnection-intents)

  • Presenter: Luis Contreras
  • Motivation: The draft addresses the evolution of network interconnection beyond simple IP traffic exchange, driven by cloud, edge computing, and virtualization. Intents are explored to leverage these new capabilities for advanced services.
  • Objectives: To enrich interconnection requests for services like deploying virtualized services in multi-domain environments or leveraging compute capabilities.
  • Applicability: Scenarios include inter-operator and private-to-public network interconnections.
  • Usage Modes: Extends beyond basic IP traffic (peering/transit) to include express service intents (e.g., CDN), VNF as a Service (e.g., packet wall), and interconnecting compute resources (e.g., FaaS).
  • Benefits: Aims to establish a common and normalized way to automate and simplify advanced interconnection requests.
  • Next Steps: Continue developing intent-based capabilities, explore specific details like policies and optical technologies, and seek community feedback. The authors also aim to position this as an IBN use case.
  • Discussion: Jérôme inquired about the primary environment for these intents (inter-carrier vs. vertical actors). Luis confirmed an initial focus on inter-operator but noted potential broader applicability through abstraction.

• Digital Twins for Industrial Networking (Project Presentation)

  • Presenter: Chiara Gracceli (University of Bologna)
  • Context: This work is part of the "FOREST" project, focusing on cybersecurity in industrial and manufacturing IoT environments.
  • Approach: Combines a Cyber Range concept with a Digital Twin. The digital twin is a virtual replica, deliberately separated from the real production environment, used for cybersecurity training, technology validation, and attack simulations.
  • Technology: Leverages Network Function Virtualization (NFV) and Software-Defined Networking (SDN), implementing the ETSI MANO framework with Open Source MANO (orchestrator) and OpenStack (VIM).
  • Lifecycle: The digital twin's lifecycle includes design (using descriptors), onboarding, automated commissioning, operation (monitoring/reconfiguration), and decommissioning.
  • Deployment Example: Demonstrated a setup including a firewall, Suricata-based Intrusion Detection System (IDS), an internal router, and a Modbus traffic generator to simulate industrial protocols (Modbus, CanOpen, MQTT, Lora).
  • Demo: A video illustrated the use of OSM's graphical interface for descriptor creation, onboarding, and automated virtual machine deployment on OpenStack. It then showed a Modbus attack simulation and the IDS detecting the behavior.
  • Discussion: Jérôme asked about the interplay between the virtual twin and the real network. Chiara explained that their current work focuses on the separated virtual environment for testing, with other project partners handling real-time interfacing. Chengsheng inquired about performance and scalability, to which Chiara responded that it depends on available hardware resources.

• Efficient Data Collection Method for Digital Twin Network Draft (draft-wang-nmrg-efficient-dt-data-collection)

  • Presenter: Wang Yang (China Mobile)
  • Problem: Digital twin networks demand real-time data from physical networks. Existing full-data collection methods are inefficient, causing issues with timeliness, storage, computation, and bandwidth.
  • Proposed Solution: A lightweight and efficient method for data collection, aggregation, and correlation. It introduces a "telemetry streaming element" (TSE) to collect data on demand, process it (aggregate, correlate, knowledge representation), and send processed, effective data to the twin network.
  • Architecture: Comprises a physical network (with data storage centers and TSEs) and a twin network (with an instruction management center and data storage center).
  • Process: Physical network components register with TSEs, which in turn register with the twin network's instruction management center. The twin network requests data, which triggers the instruction management center to configure TSEs. TSEs then collect, process (e.g., into RDF), and push the relevant data or knowledge to the twin network's data storage.
  • Advantages: Claims reduced storage resources, computing resource demands, bandwidth consumption, and data transmission latency.
  • Discussion:
    • Laurent emphasized the crucial balance between data fidelity (accuracy for digital twins) and the cost of data collection. He highlighted the need for research into optimization algorithms that dynamically determine what data to collect and when, rather than just the collection mechanisms, to align with the twin's simulation needs.
    • Chengsheng acknowledged the feedback, suggesting the next version might need to clarify or narrow the draft's scope, or expand it to address these broader aspects of data collection and fidelity.

• Digital Twin Network Concept and Reference Architecture Draft (draft-zhao-nmrg-digital-twin-network)

  • Presenter: Chengsheng Zhao (China Mobile)
  • Updates (since IETF 111): The draft now clarifies the distinctions between Digital Twin Networks (DTN) and traditional Network Management Systems (NMS), emphasizing DTN's use of interactive virtual-real mapping for closed-loop network automation. It refines the reference architecture and adds benefits related to privacy and regulatory compliance. Numerous editorial changes have been incorporated based on reviews.
  • Reference Architecture: Proposes a three-layer model: Physical Network (lowest), Network Digital Twin (intermediate, with Data Repository, Service Mapping Models, Digital Entity Management), and Network Application (top). An optional "Collection and Control" sub-layer is introduced as a southbound interface to support bi-directional data flows and processing.
  • Request: The authors requested research group adoption, noting the draft's stability after five updates.
  • Key Enabling Technologies (Appendix): The appendix outlines areas for future study: data collection (tools, semantics, aggregation), data storage & services (unified repository, fast search), data modeling (topology, functional models, AI/ML), visualization (multi-dimensional, interactive), and various interfaces.
  • Discussion: Kiran clarified if a DTN represents multiple instances or the entire physical network. Chengsheng stated it should encompass both the twin network and the physical network conceptually.
  • Chairs' Feedback: Laurent expressed support for the adoption request, viewing the document's maturity as a good foundation to identify concrete research questions for the group. Jérôme agreed, noting the "enabler" slide's utility for discussion. Chairs plan to initiate a formal call for adoption on the mailing list.

• AI Framework for Network Management Draft (draft-contreras-nmrg-ai-framework-net-mgmt)

  • Presenter: Pedro Contreras (Orange)
  • Motivation: The exponential growth of IoT and 5G networks, coupled with increasing complexity from SDN/NFV, necessitates AI-driven automation in network management. A framework is needed to ensure seamless connection, common ontologies, interfaces, and assessment methods for AI components.
  • Framework Goal: To provide a generic, scalable, deployable, and trustworthy AI framework for end-to-end network operation and management.
  • Key Components: The framework comprises a Data Framework (for acquisition, modeling, storage, distribution), AI Modules (for individual or collective data processing), and a Hub (for sharing data, knowledge, and decisions among modules and stakeholders).
  • Operation Overview: Outlines subprocesses for data collection, reasoning (context, rules), solving (building dynamic knowledge graphs to find acceptable situations), and planning (determining and enforcing actions).
  • Key Concepts: The draft introduces specific concepts for research: explicit support for closed-loop management, network intelligence (data analytics complementing AI, target-driven), external event detectors (modeling and correlating external information), anticipation of network requirements (focusing on certain future events rather than just predictions), and robust intelligent reasoning (beyond ML classification/prediction, including explanation and metadata).
  • Challenges/Next Steps: Key issues identified include defining common interfaces (including ontologies), standardizing component assessment and quality assurance, synchronizing concepts and ontologies across stakeholders, and managing information dissemination (limiting scope).
  • Discussion (Olga): Suggested integrating the intent-driven approach more explicitly into the AI framework, exploring their coexistence and mutual benefits.

Decisions and Action Items

• NMRG IBN Drafts (draft-irtf-nmrg-ibn-concepts-definitions, draft-irtf-nmrg-ibn-classification): Authors will continue to revise drafts based on IRTF chair comments and engage in planned offline discussions with Collin St. Clair.
• Research Group Adoption: For draft-dan-nmrg-network-measurement-intent, draft-lucente-nmrg-interconnection-intents, and draft-zhao-nmrg-digital-twin-network, the chairs will initiate a formal "Call for Adoption" process on the NMRG mailing list. This will allow broader community feedback on the documents' maturity, technical soundness, and alignment with the group's research direction.
• General: Authors of all presented drafts are encouraged to review the chat logs and minutes for further comments and potential references.

Next Steps

• The NMRG will continue to hold virtual meetings on a monthly basis.
• Efforts will be made to resume activities on specific topics such as "actual network and editing associates license," with a workshop being considered.
• The group aims to consolidate and integrate existing research documents.
• The mailing list remains the primary platform for ongoing discussions, draft feedback, and proposing new research topics.
• Chairs will follow up on the adoption requests by circulating the relevant drafts and initiating discussions on the mailing list as per the NMRG's established procedure for document maturity assessment.