SRv6 Operations (SRv6OPS) BoF

Friday, 22 March 2024
09:30 - 11:30 Local

• BoF Chairs: Dhruv Dhody, Thomas Graf, Weiqiang Cheng
• AD: Warren Kumari
• Slides
• Video
• Chat
• Meetecho Recording

Administrivia: Chairs (5 min)

Operator's Presentation (1 hour)

Presentations from operators highlighting potential SRv6 operational
discussion topics for the proposed WG. Each presentation had 10 mins
including time for clarifying questions.

MTN (Zoltan Miklos)

Zoltan Miklos from MTN shared the challenges and requirements of the
MTN SRv6 network.

• Warren Kumari: As someone born in South Africa, thanks for sharing
  how the networks are changing and the challenges faced.

Bell Canada (Daniel Voyer)

Daniel Voyer from Bell Canada shared the C-SID address allocation
strategy and its impact.

Swisscom (Thomas Graf)

Thomas Graf from Swisscom shared the SRv6 network observability
challenges.

• Boris Khasanov(from chat): How many vendors do support this delay
  info export in the IPFIX?
• Thomas Graf: The implementation status for on-path delay can be
  found here:
  https://datatracker.ietf.org/doc/html/draft-ietf-opsawg-ipfix-on-path-telemetry-06#name-implementation-status;
  You may also want to look at the IETF 116 hackathon presentation.

Alibaba Cloud (Linjian Song)

Linjian Song from Alibaba Cloud talked about SRv6 operations
requirement in the cloud scenario.

China Mobile (Tianji Jiang)

Tianji Jiang from China Mobile summarized the challenges in SRv6
network operations.

• Dan Voyer: Clarify if the 10K node is the total number of nodes in
  the network or just an island?
• Tianji Jiang: 10K is just in the CMNet (backbone network).
• Weiqiang Cheng: In our IP Backbone-CMNet, there are more than 10K
  core routers. In the mobile backhaul, more than 2 million routers.
  This might be the largest backbone network.

Charter Discussion (40 mins)

Dhruv Dhody went over the current proposed charter with mission, scope,
deliverables and relationships with other WGs.

• Greg Mirsky: By approving this BoF, is IESG sending a message that
  SRv6 is not IPv6?
• Warren: No.
• Greg: As per V6OPS's existing charter, it should be covered in
  V6OPS.
• Warren: This is not uncommon. If there is enough work for SRv6, it's
  ok to break it out.
• Greg: What would be the relationship between V6OPS and this new
  proposed WG is not clear to me? What is specific about SRv6 that can
  not be addressed in V6OPS?
• Warren: We could update the V6OPS charter to split the SRv6OPS to
  this new WG. Clearly, they're related but we do carve out focused
  work. Anyways point taken.
• Xipeng Xiao: V6OPS is rarely doing anything related to SRv6 right
  now. SRv6 is an application of IPv6 (not competition). SRv6OPS can
  be a useful ecosystem partner.
• Xuesong Geng: As a vendor, I support the forming of this WG for
  discussing operator presentations. Providers can give the operator's
  input that vendors/WG should incorporate.
• Tom Hill (BT): Overlap with other WGs such as NMOPS covering generic
  network management. +1 to Greg. If the WG is formed, we need to be
  critical to make sure the operations are working as well as they
  can.
• Alvaro Retna: Happy to see operators in IETF talking about the
  technology we developed in SPRING. Concerned with potential
  conflicts that we may create (ex CSID is still in development in
  SPRING). The charter text that talks about taking on work after
  SPRING should extend to others like IPPM. A lot of coordination is
  needed - overlapping chairs, overlapping ADs, everything in SPRING,
  irrespective of the charter, we need to operationalize the
  coordination.
• Andrew Alston: PCE work does not have ops wg but it works fine.
  SPRING charter gives PCEP extensions to PCE WG, BFD, TILFA, IPPM
  etc. SPRING would need to be rechartered simultaneously to avoid WG
  shopping.
• Dhruv: Yes, we need to take it into account and update impacted
  charters as well.
• Daniel Huang (ZTE): We need more operators use-cases and
  requirements; very impressive and insightful presentations. I
  believe it's quite reasonable to have focused WG.
• Cheng Li (Huawei): I see a lot of SRv6 deployment projects with
  similar problems. A dedicated SRv6 ops WG can deliver faster which
  is needed in this case. Support the WG creation.
• Éric Vyncke (from chat): With more and more IPv6 deployments, I do
  not want to have a 4-hour V6OPS meeting... Let's have focused WG
• Boris Khasanov (from chat): IMO, it's just about sharing of
  experience, that is really helpful
• Ran Pang: We deployed the SRv6 network. A dedicated SRv6OPS wg would
  be helpful for management guidance, observability, and
  trooubleshooting.
• Martin Horneffer (from chat): I do think that early exchange of
  operator consideration and discussion of best practices - including
  addressing considerations - is very helpful.
• Randy Bush (from chat): These interactions with protocol groups are
  inherent in the ops area.
• Mohamed Boucadair (from chat): Alvaro has a point but this can be
  complementary. Ops cons should be ideally handled in parallel of the
  spec, but we need some experience before making reco and guidance.
  This wg will provide more that part
• Benoît Claise (from chat): Regarding Eric's point. I could write the
  same point for NMOP: we don't to have a 4 hour NMOP meeting. It's
  interesting to observe the discussion, which is mainly "where should
  this work happen?" I assume/hope that we all agree that this work
  must happen somewhere.
• Dan Voyer (from chat): I thought it was obvious by doing this BoF
  where this work should go
• Eric Vynke: (slide 12 on scope) Refine the charter scope last
  bullet, what is the meaning of reaches consensus in SPRING?
• Robert Wilton: Good to see operators here in IETF coming together to
  solve issues. Encourage that. Could follow NMOPS in being
  operator-led and focused (not in favour of it being in the scope of
  NMOPS). Some potential overlap but solvable. Enough people and
  deployments here? It appears so. It's a good place to try. Nothing
  is set in stone, in favour of trying something here and recharter or
  closing.
• Pete Resnick (from chat): Outsider view, but seems to me that
  questions about fixing charters of other WGs, distributing work
  among WGs, which area, etc. are really up to the IESG and not for
  the BOF to decide. There do seem to be some scope concerns, which
  should be good input from the BOF, but people seem to be saying
  there is stuff to be done here.
• Robin Li: Good to see the operator here discussing SRv6 operations
  issues. Vendors had long debates in SPRING and RTGWG. Now SRv6 is
  deployed. A good platform to identify issues with interoperability.
  This WG is helpful to improve SRv6 technology.
• Jim (RTGAD hat on): We can put to bed the argument that SRv6 is not
  deployed, which is good. There is clear community interest in doing
  the work. The question is where to do it (new WG or existing V6ops
  or ...). If the WG is formed, it does not prevent us from folding it
  into v6ops later on.
• Adrian Farrel: The charter text does not show what it is not
  intended. The WG perhaps does not work on protocol extension.
• Dhruv: No standard track document.
• Mohamed Boucadair (from chat): "Solutions and protocol extensions
  are not within its scope."
• Adrian Farrel (from chat): Perhaps "Development of..."
• Dan Voyer: My experience in SPRING, BESS tells me that we do not
  have a clear voice to express the use case and challenges while we
  focus on the solutions. This new WG as a forum for operators focused
  on challenges with existing technology would be helpful. Good for
  operators and also for different vendors.
• Mahesh Jethanandani (from chat): I agree with others that NMOP is
  probably not the right place for bringing this work. I think there
  is enough focused work being defined here to form a WG.
• Yisong Liu: During the SRv6 deployment in our network, we face many
  challenges such as address planning, inter-AS, protection etc. Other
  WG may provide many technologies to solve them, but as an operator,
  we would like a place to share our experience.
• Cheng: See interest/support for the work. The agenda for SPRING and
  V6OPS is already overloaded and very hard to get a slot for
  presentation and chairs are overloaded. Why not have a more focused
  WG for that part?
• Randy Bush (from chat): this discussion kinda points out that we do
  not know how operators/operations fit in
• Antoine Fressancourt (from chat): +1 Randy, even more so since in
  other WG, I often hear struggles in finding inputs from operators
  from vendors
• Chongfeng Xie: Useful WG. The name should be reconsidered. SRv6 is
  part of IPv6 and can conflict with V6OPS. SRv6 makes the network
  complex. How to make it more reliable/secure is what we should do is
  important
• Andrew Alston (from chat): +1 - srv6 is ipv6 - or it isn't -
  creating this working group makes it explicit that it isn't - and if
  we acknowledge that then I'm quite happy with that - but lets
  acknowledge that and be open with it
• Linjian: Alibaba support the WG and would like to contribute to the
  WG. Based on Alibaba's experience on DevOps, we would like to share
  it. Seeing other operators use cases is helpful too.
• Ketan Talaulikar (from chat): From what I am hearing, there seems to
  be a need for some (or more) informational and BCP documents to be
  developed (led by operators?) alongside developing protocol work. It
  seems like creating space for that to happen at the IETF - whether
  in existing WGs or in dedicated WGs is the question.
• Greg: Agree SPRING is not the right place. V6OPS can have longer
  sessions with dedicated time for SRv6.
• Xuesong: Lots of problems can be solved without protocol extension
  such as Dan and Thomas presentation
• Jim: Need to clarify deliverables. Deliverables could be no
  documents and it's a venue for discussion. Extending SPRING to two
  sessions is an option.
• Warren: MOPS(Media OPS) provides a venue for people to share
  experiences.
• Andrew Alston (from chat): and Im not convinced there is consensus
  here - there is support - but Im not at all convinced there is
  consensus - but thats for the chairs to judge not me :)

BoF Questions

1. Are there operators here who have already deployed SRv6 in their
   networks? (Question For Operators Only)

   • 35 Yes
   • 7 No
   • 15 No Opinion

2. Are there operators here that are planning or evaluating SRv6
   deployment in their networks? (Question For Operators Only)

   • 41 Yes
   • 4 No
   • 13 No Opinion
   • Tom Hill: There might be a lot of folks working for the same
     company
   • Dhruv: Just one data point

3. Is there interest and focus in the subject matter presented today to
   warrant the formation of a WG?

   • 79 Yes
   • 15 No
   • 9 No Opinion

4. Who would be willing to contribute: Drafts, Operational
   presentations,etc?

   • 66 Yes
   • 17 No
   • 15 No Opinion
   • Greg: Is this conditional to forming WG or unconditional?

5. Who would be willing to review drafts?

   • 77 Yes
   • 12 No
   • 8 No Opinion
   • Dhruv: Never seen this many folks review a draft ever :)

6. Is there support to form a WG with the proposed charter? (Assuming
   any charter revisions will be discussed and made)

   • 73 Yes
   • 21 No
   • 14 No opinion

• Andrew: Too much overlap. SRv6 should be treated as IPv6 or ack that
  SRv6 is not IPv6.
• Greg: +1 to Andrew. a dedicated session in SPRING or V6OPS that is a
  reasonable place to have this discussion.
• Tom Hill: Not convinced with rationale. How to guarantee the charter
  change?
• Duruv: At BOF we discuss the current charter as of now, further
  changes will be made and discussed on the mailing list. This is just
  one input, it is not binding.
• Warren: The BoF question responses only indicate a direction at this
  moment in time.
• Tom Hill: Chairs would commit to charter revision based on today's
  discussion is fine.
• Matthew Bocci: Do not know what I am signing up for based on this
  BoF question
• Ketan: Existing operation consideration in the draft is never
  enough. There is a need for BCP and informational documents. This is
  beyond just SRv6 and we need a solution for routing work.
• Jeff Tantsura: On one side I love to see people presenting
  operational topics. On the other side, it is sad that this can't be
  done in SPRING. If operators need their own WG for this then
  something is broken at IETF but we should not create more
  bureaucracy. But if this is what Operators need to keep sharing
  experience, then this is good.
• Tianji: Another SDO has a dedicated venue for operation discussion.
  IETF should allow for a boost in both technology and operations.
• Warren: A SPRING session focused on OPS with separate chairs is an
  option. It is a structural problem of IETF and we need to discuss
  things.
• Xipeng Xiao: Not convinced by the argument that if SRv6OPS WG is
  formed then SRv6 is not part of IPv6. In that case, why did we
  define SRv6 in SPRING in the first place (and not in 6man and
  v6ops)?

Dhruv: Thanks everyone for a good discussion at this BoF.