NMRG 58th meeting = IETF 108, Online

RG Chairs:
* Laurent Ciavaglia
* Jéröme François

RG Secretaries
* Jéferson Campos Nobre
* Pedro Martinez-Julia

Useful links:
    * Materials: https://datatracker.ietf.org/meeting/108/session/nmrg
    * Meetecho: http://www.meetecho.com/ietf108/nmrg/
    * Notes: https://codimd.ietf.org/notes-ietf-108-nmrg
    * Audio recording: https://www.ietf.org/audio/ietf108/
    * Video recording: https://www.youtube.com/user/ietf/playlists

Wednesday 2020/07/29 11:00 - 12:40
Room 2

1. 11:00 Introduction + RG information, Chairs, 5 min

2. 11:05 Concepts of Digital Twin Network, C. Zhou, 7 min. + 3 min. Q&A
https://datatracker.ietf.org/doc/draft-zhou-nmrg-digitaltwin-network-concepts/
* Presented concepts of Digital Twin Network (DTN), benefits, challenges, next steps (see slides for details)
* Albert Cabellos: Is this technology for wired networks or other scenarios? Which kind of technology will be needed to build this system?
* Cheng Zhou: It will include both wired and wireless networks, to start with small networks in DC or telco core nets. Technology: data collection, storage, virtualization, ML, interface standardization is important to deal with different suppliers.


3. 11:15 Research Challenges in AI for NM document (discussion), J. Francois, 7 min. + 3 min. Q&A
* Reported on the status of the shared document being prepared, 7+ contributions, the core part is the challenge description, which has been categorized. The document will be used to write an I-D, first consolidated version at IETF 109. (see slides for details)
* Cheng Zhou: Use cases are in section 3, but you plan not to include them in the I-D?
* Jerome Francois: Yes, we only have a list of use cases.... [TBC]

4. 11:25 Transport Slice Intent, Luis M. Contreras, 7 min. + 3 min. Q&A
https://datatracker.ietf.org/doc/draft-contreras-nmrg-transport-slice-intent/
* Presented the draft: leveraging IBN to request network slices. Updates on draft. Link with IETF TEAS WG. Propose this an an NMRG intent use case. (see slides for details).
* Alex Clemm: Where do you see the slice controller and where its interface? What kind of intent it provides? Will it be itself an intent based system?
* Luis Contreras: Yet to be defined, it will have APIs and parameters, as an idea the IBN approach would be translated to the slice template.

5. 11:35 Interconnection Intent, Luis M. Contreras, 7 min. + 3 min. Q&A
https://datatracker.ietf.org/doc/draft-contreras-nmrg-interconnection-intents/
* Presented the draft: use of intent for interconnection. Find new models for interconnecting different operators and express it in a simple way. Leverage IBN to handle enriched interconnection requests, beyond pure IP traffic interchange. Propose this an an NMRG intent use case. (see slides for details).
* Jerome Francois: One of the next steps of NMRG work on IBN is to find use cases, so this is very important and useful. Anyone interested in IBN should look at the drafts.

6. 11:45 Service Assurance for Intent-based Networking Architecture, Benoit Claise, 7 min. + 3 min. Q&A
https://datatracker.ietf.org/doc/draft-claise-opsawg-service-assurance-architecture/
https://datatracker.ietf.org/doc/draft-claise-opsawg-service-assurance-yang/
* Presented the drafts: architecture and YANG models. Drafts were updated on Monday after feedback from Hackathon. Interesting for research: assurance graph (slide 7), which allows to visualize how symptoms affect services and where the root cause can be found. (see slides for details).
* Alex Clemm: This is a model-based reasoning to infer the faults, it typically requires a complete graph to get useful information, if it is not complete it is difficult to get the exact status, how much this model can replace monitoring?
* Benoit Claise: Monitoring today considers the network as a black box. We offer a reversed approach, we are able to tell where the problem is not looking at the specific components.
* Cheng Zhou: Is there any data? What is the accuracy to diagnose failures?
* Benoit Claise: Yes, there is data, we use also a real network. ... [TBC]

7. 11:55 Report(s) on NMRG participation to IETF 108 Hackathon,  11:55 - 40 min.
* Intent-based, assurance, telemetry, K. Edline, 15min
    * Reported on the Hackathon. Developed an open-source SAIN agent. (see slides for details).
    * Jerome Francois: You and Benoit said that the Hackathon was helpful to revise the draft. How it was actually helpful?
    * Benoit Claise: we spent one day on the Hackathon, it was very efficient. Details reported on slide 9 of my presentation.
    * Andre Bondi: Have you though about how often the probing must be done? There is a cost in doing that and this affects the response time. This is something to be carefully considered.
    * Korian Edeline: Collection every 3 sec.
    * Benoit Claise: The probe frequency depends on the service we want to monitor and the use case we are considering. It must be configurable at the subservice level. We rely on model-based telemetry, not SNMP traps.
    * Diego Lopez: we are working on mechanism on data aggregation. This can be important in your system.
    * Benoit Claise: We do have data aggregation, it's key in our system [TBC]
* A Multi-Level Approach to IBN, W. Cerroni, 15min
    * reported on the hackathon session. PoC demonstration of multi-level IBN. Very fruitful discussion with participant, although we didn't have a hands-on/coding session. First attempt to adopt the intent classification methodology. (see slides for details).
    * Olga Havel: thanks for using the intent taxonomy and classification methodology. It's good to see that no new categories were needed.
    * Walter Cerroni: it was a first attempt, we need to think about it more, but for the time being we found everything in the current classification and nothing seemed to be missing.
    * Philip Eardley: Does your approach allow multiple levels of intent (i.e. beyond the 2 levels of your hackathon)?
    * Walter Cerroni: We considered the case of physical infrastructure providers offering slice intent to VNOs or service providers and, in turn, service intents to service customers, that's why we considered only 2 levels. However, multiple levels can also be used, depending on the levels of abtractions. For instance, a complex or multi-domain infrastructure could be segmented in multiple levels of intents, e.g. one level specific to a single domain or segment (decomposition), another for network-wide needs.

* I2NSF, P. Jeong, 10min
    * Reported on the Hackathon session. I2NSF architecture. Security policy translation important for IBN. (see slides for details).
    * Charles Eckel: Do you plan to make a contribution to OpenStack to make the I2NSF framework available to the community?
    * Paul Jeong: Members of our teams are working for OpenStack, we will look into that.

Time did not allow to discuss the following extra items. The chairs will follow-up on the mailing list.
* IBN use cases and next steps (discussion)
* Intent classification (draft-irtf-nmrg-ibn-intent-classification)
* IBN Concepts and definition (draft-irtf-nmrg-ibn-concepts-definitions)
* RG infos