NMRG 63rd meeting
IETF 111, Online

Tuesday, July 27, 2021, 23:00-01:00(+1) (UTC), Room 4

RG Chairs: * Laurent Ciavaglia * Jérôme François

RG Secretaries * Jéferson Campos Nobre * Pedro Martinez-Julia

Useful links:
Materials: https://datatracker.ietf.org/meeting/111/session/nmrg
Notes: https://codimd.ietf.org/notes-ietf-111-nmrg Video recording:

Agenda + Notes

23:00 Introduction, RG Chairs, 5 min.

23:05-23:15 - Concepts of Digital Twin Network, Cheng Zhou, 5 min. + 5 min

Laurent: in the architecture layered view, the control (actions on the physical network) always directly goes through the DT, can we have also direct control between the app and the physical network and the DT used in parallel?
Cheng: not well addressed yet but the applications running of the physical network could be also a represented objec. We will consider it in next versions.

23:15 - 23:35 - A practical application of monitoring data and AI in supporting
an industrial environment (a robotic digital twin), Diego Lopez, Ignacio
Dominguez, Daniel Gonzalez, 15 min + 5 min (Q&A)

Jerome: for the configuration of monitoring jobs and data flows, does this depend only on the model to be instantiated or can it be also adjusted based on some user requirements (for example, expected accuracy) ?

-Diego: it can be adjusted regarding many parameters, is more related to the presentation "Towards Building a Context-Aware Data Aggregation Framework for Network Monitoring"

Yong-Geun: It is very interesting. I am wondering whether you have any goals or degrees (e.g., reduce delay time or increase accuracy) compared with existing implementations.

Pedro: Have you estimated the actual amount of memory and CPU resources required to build a high-fidelity digital twin of a network?
-Diego: here it is a specifc DT for our robotic use case
-Cheng: We didn't not evaluate the resource consumption yet. Basically it depends. DTN can be built phase by phase, and from inital small network domains, specific topology, to big domain with whole topology.

23:35 - 23:50 - An Efficient Data collection method for Digital Twin Network,
Danyang Chen, 10 min + 5 min (Q&A)

Jerome: shifting the problem from the DT to the real network
- Danyang: will consider the suggestion in future work

Laurent: is there an ontology developed/maintained? if yes, by whom, what is the ontology and how applied to heterogeneous/new technologies?
Ignacio: an ontology would be ideal. As we integrate a new data source, we try to capture all context information

Laurent: form of the final publication? in a journal paper or as an RFC? Also, do you consider different versions with sets of challenges?
Jerome: no preference has been expressed by contributors who answered this question (also raised several times)

00:25-00:35 - Network measurement intent, Danyang Chen or Kehan Yao, 6 min + 4
min Q&A

Laurent: good progress of the work, and indeed could be interesting to consider the work in relation to the IBN use cases, or in relation to the IBN classification document

00:35-00:45 - Status of RG documents, RG chairs, 5min + 5min Q&A

Colin: timeframe for submitting
Laurent/Jerome: documents are stable and do not requires work from the group. Need some syncrhonization with the authors for last edits of the documents to be ready for IRSG review (next couple of weeks)

Reinforcement Learning-Based Virtual Network Embedding: Problem Statement,
Ihsan Ullah https://datatracker.ietf.org/doc/draft-ihsan-nmrg-rl-vne-ps/