nmrg

Summary

The Network Management Research Group (NMRG) session at IETF 117 focused on network digital twin (NDT) research. The session included a presentation on data collection requirements and technologies for NDT, followed by an open discussion about defining a research agenda for the group. Key topics included use cases, security and privacy considerations, the layered approach to NDT, and the importance of research questions.

Key Discussion Points

Data Collection Requirements: Discussed various data collection requirements for building digital twin network systems, including topic-driven collection, diverse tools, lightweight collection, open standardized interfaces, naming conventions, and efficient destination delivery.
Distinguishing Data for NDT vs. AI: Clarified the distinction between data collected for NDT and general AI applications, emphasizing the comprehensive and real-time nature of NDT data requirements.
Scope of NDT: Discussed whether the goal is to simply collect network information or to replicate network equipment in the digital twin. Need to clarify level of replication fidelity desired.
Research Challenges: Identified the need to focus on research challenges rather than engineering tasks in the context of NDT. How can predictive AI methods steer data collection to validate or correct predictions?
Network Telemetry Subscriptions: Raised the issue that network telemetry subscriptions should be adjusted based on service deployments and aligned with data mesh architecture.
Use Cases: Emphasized the importance of defining specific use cases for NDT, such as verifying AI solutions, training AI models, and creating sandbox environments.
Security and Privacy: Addressed security and privacy aspects, including secure access to information, data reliability, and trust issues.
Layered Approach to NDT: Explored the concept of a layered approach to NDT, considering different network layers (IP, Layer 2, optical) and the need for varying resolutions.
Data Plane Scaling: Discussed challenges related to data plane scaling in NDT, particularly resource utilization and performance under different loads.
Model Abstraction: Raised the question of how coarse a model can be while still being useful for specific purposes, balancing abstraction with accuracy.

Decisions and Action Items

Action Item: Collect comments and additional feedback to aggregate them and build a research agenda proposal. (Jean Michel)
Action Item: Animate definition of concrete definition with text, milestones, something organizing a bit this this approach. (Jean Michel and Jerome)

Next Steps

Formulate good research questions on the topic of NDT and set objectives for proposed work.
Articulate the different propositions together to understand relationships between data collection, data generation, and modeling.
Concentrate efforts on priority topics/focused areas to generate momentum and progress.
Consider the form of results, including Internet drafts, publications, implementation proof-of-concepts, and demos.
Open and liaise with other groups and communities inside and outside of the IRTF and IETF.
Initiate work on network digital twin research challenges.
Concentrate on design guidelines and architecture.
Highlight specific use cases.
Consider standard implications.

Session Date/Time: 28 Jul 2023 00:00

nmrg

Summary

The Network Management Research Group (nmrg) met to discuss key research topics including Green Networking, Intent-Based Networking (IBN), Data Management, and Artificial Intelligence (AI). The session reviewed current efforts, solicited feedback, and discussed future directions for research within the group. The chairs emphasized the importance of focusing research efforts and connecting with other relevant groups and SDOs.

Key Discussion Points

Green Networking:
- Challenges and opportunities were presented, organized around device/equipment, network-wide, protocol-level, and architectural aspects.
- Topics included carbon telemetry, network optimization using AI, and energy-saving mechanisms.
- Discussion emphasized the need to focus on longer-term research challenges and ensure short-term standardization efforts are addressed within the IETF. The need to coordinate with the AAV Impact program was also highlighted.
Intent-Based Networking (IBN):
- There was a willingness to have more things regarding IBN in the group.
- The importance of focusing on use cases was noted, as well as progressing beyond the current short-term agenda to include architecture, mediation, assurance, interfaces, intent representation, and modeling.
Data Management:
- Data fabric and data mesh paradigms were presented as approaches to unify and manage data from heterogeneous sources.
- Knowledge graphs were discussed as enablers of these paradigms, leveraging semantic web technologies and property graphs.
- Challenges include mapping different data models (e.g., YANG) to common concepts and the need for network industry ontologies.
- Concerns were raised about the complexity and potential overhead of transforming existing data into ontologies. Alternate approaches were discussed, such as building ontologies on top of existing YANG models.
- The importance of considering the energy footprint of data management methods (e.g., digital twins) was raised.
Artificial Intelligence (AI):
- Discussion focused on challenges documented in a previous AI challenges document.
- The need for use cases to drive the exploration of challenges was emphasized.
- The relationship between AI and Machine Learning (ML) was discussed. It was suggested that AI should be viewed as broader than just ML.
General Remarks:
- The group was encouraged to articulate the research questions being addressed.
- Various forms of contribution were encouraged, including drafts, presentations, demos, and participation in hackathons.
- Stronger connections with other IETF groups, other research groups, and SDOs were encouraged.
- The importance of focusing research efforts and concluding ongoing work was emphasized.

Decisions and Action Items

Green Networking: Hold an interim meeting to determine a possible agenda.
Intent-Based Networking (IBN): Encourage contributions, particularly regarding use cases. If no substantial contributions are received by IETF 119, consider ending the IBN topic.
Artificial Intelligence (AI): Hold an interim meeting to redefine the challenges to focus efforts on.
Data Management: Hold a dedicated meeting to focus efforts on specific topics.
The chairs will work to organize more frequent interim meetings and discussions between main sessions.
Participants are encouraged to explore collaboration opportunities between different topics.

Next Steps

Schedule interim meetings for Green Networking, IBN, AI, and Data Management.
Solicit contributions from participants, particularly use cases for IBN.
Evaluate the scope of ongoing work and prioritize efforts to ensure timely completion.