Minutes IETF101: coms
minutes-101-coms-00

Common Operations and Management on network Slices (COMS)

Meeting Minutes

COMS BoF @ IETF-101, London

09.30 - 12.00, Thursday March 22nd, 2018

Chairs:

- Gonzalo Camarillo
- Adrian Farrel

Agenda: https://datatracker.ietf.org/meeting/101/agenda/coms
Jabber room: coms@jabber.ietf.org
Mailing List: https://www.ietf.org/mailman/listinfo/netslices
Meeting materials: http://tools.ietf.org/agenda/101/

Minutes Takers:
Ignacio and Christian

1. Administrivia - chairs [2 mins : 2/150]
Slides: https://datatracker.ietf.org/meeting/101/materials/
                  slides-101-coms-chairs-opening-slides-00

2. Objectives of the BoF - ADs [3 mins : 5/150]
- Ignas Bagdonas (OPS&MAN AD): on the purposes of this BoF
- Intro comments from Jeff Tantsura (IAB)

3. Overview of Questions to be Answered - chairs [5 mins : 10/150]
Refresher on the questions the ADs want answered.
See item 8 on the agenda.

4. Setting the Scene - Jari Arkko [20 mins : 30/150]
Reference: https://tools.ietf.org/html/draft-arkko-arch-virtualization/
Slides: https://datatracker.ietf.org/meeting/101/materials/
                  slides-101-coms-chairs-opening-slides-00

 * We have already a lot of tools
 * Do we need something new?
 * Stark contrast: device config vs. orchestration
 * Find concrete missing things (after considering _all the technology_)
 * _General over specific_ (not just 5G)

Q&A

  Reza Rokui (Nokia): NS is end-to-end.
     It is critical to address it: RAN, core and application.

  Gonzalo: Let's hold this question until the Q&A

  Reza: Yes, but we need clarity (terminology matters).
     What do we want to produce?

  Pat Thaler (Broadcom): Consider collaborating with DETNET

  Hannu Flink (Nokia): On the 5G scope

  Jari: Should be generic but allowing tools for customization
     (e.g., specific for 5G)

  Hannu: How do you relate top-down and bottom up?

  Jari: Yeah, you need to sum them up, e.g. top-down (e.g., models)
     together with bottom-up (e.g. tools)

  Gonzalo: Both approaches to be discussed as part of item 8 "The
     Questions" discussion in the agenda.

5. Why are we trying to provide NS solutions [30 mins : 60/150]

a. Motivation for Management of Network Slicing and IETF work from
   Operator's View Point -  Shunsuke Homma (NTT)  [10 mins]

   Slides: https://datatracker.ietf.org/meeting/101/materials/
      slides-101-coms-motivation-for-management-of-network-slicing-
        and-ietf-work-from-operators-view-point-shunsuke-homma-01

   - Proposal for COMS work scope (slide #9) and COMS Deliverables:
     (i) Information/data modeling,
     (ii) Interfaces for Interworking and Stitching,
     (iii) Clarification of data plane functionalities and how to
           configure them

b. Motivation for Management of Network Slicing and IETF COMS work from
   Operator's View Point -  Luis M. Contreras (Telefonica) [10 mins]

   Slides: https://datatracker.ietf.org/meeting/101/materials/
      slides-101-coms-motivation-for-management-of-network-slicing-
      and-ietf-coms-work-from-operators-view-point-luis-m-contreras-00

   - Highlights the multi-domain problem (interconnections with other
     providers and among affiliates) fo provide services outside the
     footprint of a primary provider

   - Challenges faced:
     - Strict SLAs
     - High customization
     - Traffic segregation (physical/overlay/slicing?)

   - Call for new interconnection model --> How do you communicate
     among different service providers?

     --- Standard procedures required to minimize integration costs of
         distinct administrative domains)

Reference:
    draft-qiang-coms-use-cases

Supporting references:
    draft-geng-coms-architecture
    draft-homma-coms-slice-gateway
    draft-defoy-coms-subnet-interconnection-03,

c. Q&A  [10 mins]

Reza Rokui (Nokia): RAN and core is missing (referring to the slides in
   the previous presentation)

Parviz Yegani (Huawei): A bigger picture is missing, connection between
   administrative domains is much more complex tham just connectivity at
   multi-domain (cites related work at MEF LSO).

Adian (chair): Are you asking a question?

Parviz Yegani: I want to expand this view on and end-to-end prespective.
   E.g., information model is missing. Can IETF help with it (i.e.,
   unifying information models) beyond data models (e.g.good pieces of
   work like L2SM, L3SM)?

Gonzalo (chair): This is a larger discussion, let's hold it for the Q&A
   later

6.  What are we trying to do?  [40 mins: 100/150]

a. Problem Statement and Architecture of COMS
   Liang Geng (China Mobile) [10 mins]

   Slides: https://datatracker.ietf.org/meeting/101/materials/
      slides-101-coms-problem-statement-and-architecture-of-coms-
      liang-geng-00

   - Focus on common operation and management with mapping to underlay
     technology doamins but not data plane

b. COMS Architectural Design Enablers & Artefacts (1) :
   COMS Technology Independent Information Model -
   Cristina Qiang (Huawei) [10 mins]

   Slides: https://datatracker.ietf.org/meeting/101/materials/
      slides-101-coms-coms-architectural-design-enablers-artefacts-1-
      coms-technology-independent-information-model-cristina-qiang-01

c. COMS Architectural Design Enablers & Artefacts (2) :
   Network Slice Interconnection -
   Xavier de Foy (InterDigital) [10 mins]

   Slides: https://datatracker.ietf.org/meeting/101/materials/
      slides-101-coms-coms-architectural-design-enablers-artefacts-
      2-network-slice-interconnection-xavier-de-foy-03

Reference:
    draft-geng-coms-architecture

Supporting references:
    draft-defoy-coms-subnet-interconnection
    draft-homma-coms-slice-gateway
    draft-qiang-coms-netslicing-information-model

d. Q&A [10 mins]

Reza Rokui (Nokia): The information model should NOT have. ???.
   Multi-tenancy is missing. Which are the components inside the NSO to
   resolve a request?

Liang: About NSO: we had a big diagram with NSO elements, but it is
   implementation specific. Check the operational guidance, but we need
   to discuss it.

Reza Rokui (Nokia): What will provide IETF (inside the NSO)? We have to
   understand what IETF is going to provide.

Gonzalo: This is one of the key questions to be discussed/answered

???(Tsinghua university): Do you need to unify the SPI?

Liang: I do not think is practical. We previously discussed it.

???: Yes, there is a concept of multi tenancy. About storage and
   compute for NF: you need to know it so you can allocate resources
   for it

Reza Rokui (Nokia): I agree but the info should not be inside that.

???: That's what I'm trying to say.

Myung-ki( ETRI): Is the radio resource managment inside COMS's scope?

Cristina Qiang (Huawei): No, because they belong to 3GPP

Liang: The NSO may need East-West interface, but the 3GPP is doing it.
   Some day we'll need a guaranteed transport to isolate ran from COMS.

Raphael Rosa (University of Campinas): About multi-domain information
   model: are you considering addressing different degrees of
   abstractions views for a domain  exposing and requiring
   resources/capabilities this in the draft?

Cristina Qiang: Yes, we want to have a unified view for the NS. But it's
   hard to stitch them togethter

Raphael: Are you considering different degrees of freedom  (e.g.,
   hierarchy, recursion)?

Liang: You can always play with attributes.

Chunsan Xiong (Huawei): If we want to open this interface to third
   parties, FACPS needs to be openly defined in information model and
   interface to third party and cross different administritive domains.

???: We listed it out as one of the requirements

Chunsan Xiong (Huawei): From my undertanding, we should address also
   perfomance and security issues.

Warren Kumari (Google): I don't really get the house analogy. There's a
  lot of discussions when building "houses". Can I ask for a quick poll?
  People who work for vendor /NS slice tennant can raise their hands?
  (afterwards) People who would be willing to be NS providers can raise
  their hands?

7.  How do the concepts fit together? [15 mins : 115/150]

a.  Alex Galis (UCL) [10 mins]

    Slides: https://datatracker.ietf.org/meeting/101/materials/
        slides-101-coms-how-do-the-concepts-fit-together-alex-galis-01

Reference: draft-geng-coms-architecture

b. Q&A [5 mins]

   No questions

Ignas Bagdonas (Equinix):
   How many people would work on operations in this context?
   (counted less than 10 hands)

   Provider of a Service or Broker of a Service?
   (counted 6 identified as service providers??;
    counted 2 identified as Brokers??)

8. The Questions - chairs [25 mins : 140/150]
   Move through the questions one-by-one getting clarity on the answers.

   Jeff Tantsura (Nuage networks): We have to address the definition of
      a slice if we want to see progress. Also security. What will you
      measure to verify SLA compliance?

   Adrian (chair): Who in the room understands what the proponents want
      to achieve? Approx (well over?) 40 hands up.

      Do we need to address all the presented issues day 1?
      Or let's do single domain first and then to the bigger picture?
      Poll:

      - Full picture (end-to-end)? 50
      - Start with single-domain? (edge-to-edge) 10-12 hands

   ?? We have overloaded the term domain (and the term service??).

   Jari Arko (Ericsson): I think we made a lot of progress today. Are we
      trying to describe globally what the network does? >> too broad in
      my opinion. IETF is very good doing data models, but we may be not
      so good at the orchestration level.

   Warren Kumari (Google): Asks whether answering to the full picture vs
      single-domain scope question implicitly implies agreeing to the
      work, both questions should start with "IF we do this",

   Alex Galis (UCL): you cannot move easily from single (now) to multi-
      domain (in a 2 years). Part of the multi-domain protocols need to
      be attacked now. Multi-domain is not orders of magniture more
      complicated than single domain, as proven by some recent
      (research) prototyping efforts

   Reza Rokui (Nokia): The single-multi domain issue depends on who you
      ask: customer or provider. Note that you have to  address the
      end-to end issue. We need to suport a service

   Charles Eckel (Cisco Systems): There is work being done in other
      places, check the gaps that IETF can fill.

   Adrian (chair):
      Does anybody in the room have a strong disagreement with the
         architecure picture?
      I see no hands up.

      About interfaces:
         Did anyone come away from the meeting unclear about the CSI
         and SDI?

            Parviz Yegani (Huawei): What do you mean with interface?

            Adrian: internal: API, external: protocol

            Jari Arko (Ericsson): I think the architecure is right but
               not how much information about the network we are
               specifying

            Lorand (jabber room): Operation and management of the slice
               is missing.

            Young Lee (Huawei): SDI is internal or external? There is
               some ambiguity in the SDI definition as/if they could be
               recursive regarding CSI

   Adrian: Which are the must-have deliverables?

   Adrian: Relation with other SDOs: we've heard very presentations
      about that. Does anybody think there is work in IETF that fits
      COMS' work? I see ???

   Adrian: Top-down or bottom up: not enough clarity yet. IETF typically
      is better at bottom up but should be discussed here.
      - Who wants top-down? 30
      - Who wants bottom-up? 30
      - Open source? Fewer than 10

      Should we work with other bodies? ?? Which?
         Shouts from the audience: ETSI, 3GPP, MEF
         Dave Sinicrope notes that there are very many SDOs working in
            this area

      Who thinks the IETF should own the full architectural picture?
         No hands

9. Wrap up - ADs [10 mins : 150/150]

   Ignas Bagdonas: Will the energy for work sustain the future?
      Who is willing to work on that? Who is willing to coordinate
      with other SDOs?

   Alex Galis (UCL): Right now there is no need to coordinate with other
      SDOs, just orchestrate/interoperability. Only IETF can deliver
      the ??? Should be make compatible to make easy to adopt for the
      industry.

   Ignas Bagdonas: unclear if we will create/modify protocols. The
      aforementioned interoperability may be impractical. The problem
      scope seems to be finally clear. The main group of authors seems
      not open enought to global IETF public. At this point of time it
      seems there is energy on getting work done

   Alex Galis (UCL): My personal view: if no group of cohesive drafts
      are put together, the energy will disappear. We should discuss the
      group scope but there is limited time. We clearly need to move on
      to make sure it is used. Recall that there are at least 25 years
      of research in network slicing.

   Kiran Mahanhagi (Huawei): Do you plan to spread the work on other
     SDOs?

   Ignas: ???

   Liang (China Mobile): I think there's a pretty good ratio of raised
      hands taking into account the number people who raise hands in
      IETF. I think we have a pretty good understanding of the scope
      and energy. We should take action now (not necesarily forming a
      WG).

   Ignas: Way forward: consult with people/customer. This needs to be
      socialized outside of the IETF.

   Liang (China Mobile): 6 hands is good taking into account the number
      of operators in IETF

   Ignas: the work does not stop if there is not WG, self-organize for
      a BoF. Try to reach more to the operational communities.

   Gonazlo: adjourn