Computing-Aware Traffic Steering (cats) WG Virtual Interim Meeting

• When: Wed 15 Jan 2025, 13:00-15:00 GMT
• Co-Chairs: Adrian Farrel, Mohamed Boucadair, Peng Liu
• Minutes: Cheng Li, Adrian, Med, Dirk

1. Introduction & Administrivia

Agenda

1. Settling time (5 mins, 5/120)
2. Administrivia - chairs (5 mins, 10/120)
3. Use cases and requirements
   2a. draft-jeong-cats-its-use-cases - Jaehoon (Paul) or Bien Aime (15
   minutes, 25/120)
   2b. draft-ftzhs-cats-industrial-requirement - Fu Tao (15 minutes,
   40/120)
   2c. draft-lcmw-cats-midhaul - Luis (15 minutes, 55/120)
   2d. draft-wang-cats-green-challenges and
   draft-wang-cats-usecase-green - Jing Wang (25 minutes, 80/120)
   2e. General discussion of use-cases and requirements with next steps
   • All (10 minutes, 90/120)
4. Metrics draft-ietf-cats-metric-definition - authors (15 minutes,
   105/120)
   Open issues and work plan
5. AOB for the WG - All (15 minutes, 120/120)

[Med] Today's meeting will focus on the use case draft and metrics
draft. Regaring use cases draft, we will have multiple presentations, to
see if we can merge the use case into the main use cases draft, abandon
the individual I-D, etc. Regarding the main use cases draft, we may
discuss whether we(WG) want to publish the draft as an RFC.

[Adrian] Our Infomational use cases draft does not need to include ALL
use cases, but only need to select some typical use cases. No need to
prove anything by adding use cases, because we have a WG already.
Today's goal is to discuss if a use case is within the charter, and do
we want to merge it into the main draft. The chairs would like to try to
finalize the content of the WG draft getting consensus on its content
and freezing it or asking for publication.

Authors of draft-bernardos-cats-anchoring-service-mobility replied but
are not ready to present this time. Authors of
draft-jiang-cats-usecase-5gedge didn't replied, so we will delay the
discussion to next meeting in Bangkok.

2. Use cases and Requirements

Context & Goals:

• Recall that this document paints a picture of some key use cases. It
  is not intended to be a comprehensive list of all use cases that
  will ever be applicable.
  Further, it is not a "proof of necessity" for CATS -- we already
  have a working group!
• Our objective should be to get the working group use-case draft
  ready for a consensus call as "complete", with that call to be held
  immediately after Bangkok
• Does the WG want to push for RFC publication, or just tag the I-D as
  having consensus
• For all other use-case-related I-Ds:
  • Are they already included in the WG use-case draft? If so, can
    we "close" the other draft?
  • Should they be included in the WG use-case draft as main use
    cases or as "future uses" for the appendix?

2a. draft-jeong-cats-its-use-cases

[Adrian] Helpful presentation. Two actions:

1. Look at the existing use case draft, section 4.2, suggest some edits
   to the text already in the draft. You can contribute some text to
   section 5 as well. Please suggest change/text (don't wait for the
   authors of the WG draft to do it).
2. Change this draft, removing use case/reqs, change it to an
   applicability draft. It may be a slow work, because you may need
   framework and solution work to progress as well.

[Paul] Is it possible to add an appendix?

[Adrian] Prefer to see you propose some text in the main body of the
draft.

[Med] We only need the core use cases, not all, so please contribute
to the use case draft directly.

[Paul] Comfirmed again: will contribute to section 4.2 in the use
case.

[Shihang] Agree with Chairs. Please use github to raise a PR. A
question, it seems there are some solution related text in the draft,
like protocol related text. We may not need this in the use cases and
requirement draft, please handle this.

[Paul] We will handle that.

2b. draft-ftzhs-cats-industrial-requirement

[Futao] Will try to contribute some text to the existing use case
draft. Will study the security issues in industrial scenarios.

[Med] Thanks. @Tao, please focus on new requirements and summarize the
UC.

[Adrian] Please prepare two pages of text to the use case draft, we
can discuss it on the mailing list to see if we can add it into the use
case draft as a new subsection of section 4. Also look at the
requirements in section 5 and see if any work is needed.

[Futao] Will do.

[Peng] Please also addree the comments from the past meetings. Try to
present your work in Detnet, or you can only focus on the CATS
requirements and use cases in CATS working group.

[Futao] Will do.

2c. draft-lcmw-cats-midhaul

[Med] On your first point, I do not see new specific requirements from
this draft.

[Luis] A specific requirement from this use case is the connection
with external systems. current CATS UCs all assumes that all components
are internal

[Med] Can you share this on the mail list? regarding the second point,
how should we proceed? Should we consider an LS to O-RAN or proceed in
the other way around O-RAN to the IETF?

[Luis] Will do, and will try to trigger some discussions in metric
draft.

[Shihang] O-RAN is controvesal and we are not expert of O-RAN, it may
be premature to discuss O-RAN. It may be good to have a liasion from
O-RAN then we can start the work.
[Luis] Thanks. We are focusing on Midhaul, it can apply to O-RAN and
3GPP. O-RAN is not for one signle RAN, but for 4G, 5G and now 6G.

[Shihang] O-RAN is only one kind of RAN, there are other RANs now. We
may need to discuss all RANs instead of only for O-RAN. If O-RAN needs
CATS, please send a Liasion.

[Tianji] As co-author, O-RAN is useful. Liasion from ORAN will be
helpful on promoting the draft. it is useful to have it in use case&req
draft, while it does introduce extra requirements.

[Med] Luis, do you agree that we should ask for a Liaison?

[Luis] Let me discuss with Mark, may come back with some ideas.

[Med] Please go ahead, I like to go with this way. Thanks.

2d. draft-wang-cats-green-challenges and draft-wang-cats-usecase-green

[Cheng from chatbox] Energy related metrics are important in selecting
a cheaper(may be better) serivce contact instance. We can add it as a L1
metric in metric definition draft and it can be considered as an
influencing factor in generating the normalized L2 metric, see
discussion:
https://github.com/ietf-wg-cats/draft-ietf-cats-metric-definition/issues/5

[Shihang from chatbox] Engergy related metric can be added in to the
metric draft.

[Dirk from chatbox] Agree with adding energy as a L1 metric, which can
then feed into the normalisation at L2 level -> you get EATS. It will be
good to have it as a use case in use case draft.

[Adrian] Do you want to discuss the energy not only on the computing
side but also on the network side?

[Wang Jing] Yes.

[Adrian] We may have some overlaps with other WGs regarding the
network energy part. So I do not want us to go too far ahead. I may
prefer to make it as a future applicability thing, instead of doing it
on day one.

[Dirk] It is worthy to add costs as a metric for CATS. For example
adding as a L1 metric.(similar to the text in the chatbox), which then
feeds into the overall normalised L2 metric. This allows for considering
costs as a decision criteria as part of CATS.

[Med] Please continue the discussion, but I am not sure it is good to
have it in the use case draft or not. There are some considerations that
are more an applicability statement, but I'm afraid we don't have text
to back the cost discussion let alone that cost is not specific to
energy.

[Dirk] Given that a discussion on (green/energy) costs informs the
(requirements for the) metrics draft, I tend to see this as part of use
cases, while before 'applicability' may have been enough.

[Med] Do we have any specific discussion on energy and cost?
Please go to the use case, requirement of the draft, to see what
specific requirements are needed, and share with the mailing list to
proceed the discussion.

[Dirk] I will see how to add something, possibly to an existing use
case, for consideration of costs, which then leads to an additional
requirement without really adding specific use cases.

[Adrian] We need to do some homework. It may impact more on the
requirements than use cases.

[Julien Maisonneuve] A specific use case of energy could be necessary.
Also new requirements, and metrics. We already have too many use cases
now, should prioritize them based on the specific requirements they
bring to ensure we address them.

2e. General discussion of use-cases and requirements with next steps

[Adrian] The authors of the presented drafts, please contribute text
to the use case proactively. Please do not wait or ask the authors of
the use cases draft to do this for you. Thanks.

[Kehan] Agree with merging the usecases drafts into the main one.
Energy ralted discussion is useful, it will generate some requirements.
But we need to thingk carefully, especially the implementation part.

3. Metrics

Goals:

• Follow up on the adoption poll discussion
• Decide on next steps
• Can we start to quantify the amount of information and how often it
  varies?

draft-ietf-cats-metric-definition

[Kehan] Authos agree with option B(Covering all the computing side
metrics). Will provide a formal reply before next update.

[Med] On the storage part, do we have a case to illustrate how
storage-related metrics are key for instance selection?

[Hang] Different instance may be bottlenecked by different part of the
system. CPU/Storage/network are all useful to measure the load of the
instance.

[Hang] I think CDN is one usecase that may be bottlenecked by storage.

[Kehan] If we can consider storage IO speed as storage metrics, I
think it's useful for CDN use case in the appendix. But it's not well
backed by the main body use cases of the draft.

[Luis] Agree with Hang and Kehan that storage is relevant and CDN
showcase that relevance

[Med] This should be discussed in the drafts to motivate that use.

[Adrian] Look forward to having the next revision before Bangkok.

4. AOB for the WG - All

Open Mic questions

[Tianji] Already 3 WG documents, but how about other contributions?

[Adrian] Reasonable question. We may be ok to start to look at the
solution. But we may need to get the ground work done firstly. But since
we have understood what are the metrics, it may be ok to see how to
distribute the metrics (solutions).
If we look towards the botom of the charter, we see "Additional
groundwork to include:" and "Applicability of existing tools and
mechanisms:" There are topics here that we can now start to investigate.

[Med] Please review the existing documents, and contribute to the
documents, show evidence that we have confidence to reach the existing
milestones in time. Then we can look at new work.

[Chairs] See you in Bangkok.