NMRG 67th meeting
IETF 115, Hybrid
Monday 2022-11-07 09:30 UTC
https://datatracker.ietf.org/meeting/115/session/29826.ics
Contacts:
RG Chairs
RG Secretaries
Useful links:
Agenda:
09:30 Introduction and RG status, RG Chairs, 05 min
Intent Based Networking
Artificial Intelligence for Network Management
Network Digital Twin
Miscellaneous
Notes:
09:30 Introduction and RG status, RG Chairs, 05 min
RG updates:
Intent Based Networking
IETF Network Slice Intent, Luis Contreras, 10 min. / 09:35
https://datatracker.ietf.org/doc/draft-contreras-nmrg-transport-slice-intent
Q-Mohamed Boucadair: On Slide 3: is tagging an intent as SLO or
SLE part of the intent "formulation" itself or is it part of the
translation process?
A-Luis: part of the translation process
Q-Alex Clemm: what is subject to standardization vs. research?
A-Luis: intent expression, assurance phase potentially more
research topics. translation process rahter more straightforward
and for PoC / standardization (TEAS?)
Q-Olga: Do you use TMF intent model or do you start from
scratch?
A-Luis: offer that in a general manner
Interconnection Intents, Luis Contreras, 10 min. / 09:45
https://datatracker.ietf.org/doc/draft-contreras-nmrg-interconnection-intents
Q-Dhruv: is it customer-provider or provider-provider
interactions?
A-Luis: provider to provider
Comment from Jérôme on question from authors
Two options for progressing the work
1. For authors of one draft to join the effort of the other
draft, working together on separated drafts
2. To merge the drafts while maintaining differentiated
sections for the specificities of each draft
• Question to the Chairs and the RG: What do you think is
the best approach?
Jérôme: need first to identify commonalities to see how
much could be merged and if this would make sense.
Network measurement intent, Kehan Yao, 10 min. / 09:55
https://datatracker.ietf.org/doc/draft-yang-nmrg-network-measurement-intent
Discussion on approach for IBN use case documents: a common
document or separate ones.
Previous poll on preference - outcome ?
-> suggestion: bring the discussion to the mailing list and
authors/RG participants to express preference.
Artificial Intelligence for Network Management
Research Challenges in Artificial Intelligence for Network
Management, Jérôme François, 10 min. / 10:05
https://datatracker.ietf.org/doc/draft-francois-nmrg-ai-challenges
Q-Diego: sustainaiblity of supervised method? what about
distributed AI (not only federated)? maybe to add
A-Jérôme: looking for some help regarding federated/distributed
AI
A-Diego: I can help on that
Comment-Laurent: important to "publish" the challenges (RFC or
article); and next step: investigate one or more challenges in
NMRG (and with ohter groups / communities)
Considerations of deploying AI services in a distributed
approach, Yong-Geun Hong, 10 min. / 10:15
https://datatracker.ietf.org/doc/draft-hong-nmrg-ai-deploy
Q-Alex: more system management or network management?
considerations for federated learning, transfer learning? would
like to see more "network-side"
A-Yong-Geun: agree. welcome inputs.
Q-Jeff: mentioned "vertical" hierarchy (remote/close edge) vs.
"horizontal" hierarchy (ability to run the inference on one or
multiple servers)
A-Yong-Geun: agree. part of our study. will put more details on
this consideration
Comment-Jeferson: It would be nice to have the similarities
between this draft and the research challenges one pointed out
Network Digital Twin
Performance-Oriented Digital Twins for Packet and Optical
Networks, Chris Janz, 10 min. / 10:25
https://datatracker.ietf.org/doc/draft-paillisse-nmrg-performance-digital-twin
Q-Sharon Barkai: consider making DT of the traffic patterns?
(IPFIX-based twins)
A-Chris: consdered in the first version. not relevant for the
optical network DT.
A-Jordi: not currently in scope.
Q-Laurent: Do you consider also "configuration" type of
interactions on the NII interface? e.g. to configure
collection/measurement/monitoring information flows to the PODT?
(or are these configurations handled at another interface / by
other processes)
Q-Alex: wrt. question on slide #7 (Information-DT and/or
Action-DT). consider Action-DT is the more general case.
A-Chris: new capability provided by information-DT. Action-DT,
management system can be built using other/known constructs
(intent renderer, closed loops, etc.)
A-Jordi: definition of NDT. scope of Action-DT too broad.
Requirements for Interfaces of Network Digital Twin, Danyang
Chen, 10 min. / 10:35
https://datatracker.ietf.org/doc/draft-paillisse-nmrg-performance-digital-twin
Comment-Laurent: need discussion on interfaces to network
digital twin (cf. also previous draft/presentation)
Comment-Jérôme: new I-D. Invite comments on the mailing list.
Miscellaneous
Challenges and Opportunities in Green Networking, Alex Clemm, 15
min. / 10:45
https://datatracker.ietf.org/doc/draft-cx-green-ps
Q-Jeff: reaching limits. cannot deploy anymore due to amount of
energy required.
Q-Diego: in support. taking into consideration... similar to
previous discussion by Chris(?) regarding information vs. action
Q-? Clarification of scope? Tests to prove the sustentability
of a network
A-Alex: holistic view vs "network" view can be considered
Q-?: Interesting topic. Just energy consumption or other issues
such as recycling?
A-Alex: Some metrics (sustentability metric?) could consider
other issues.
Comment-Jeff: Power consumption metric could be considered for
computation
Comment-Laurent: it is important to frame the problem space and
what NMRG can realistically address
Modelling Boundaries, Nigel Davis, 15 min. / 11:00
https://datatracker.ietf.org/doc/draft-davis-netmod-modelling-boundaries
Comment-Laurent: it is an interesting topic. would be good to
continue the dsicussion*