Notes-IETF-116-cats (Yokohama and Online)

Wednesday 29 March 2023
Room: G403
15:30-17:00 Tokyo

See Materials: https://datatracker.ietf.org/meeting/116/session/cats
Note taking: https://notes.ietf.org/notes-ietf-116-cats
Meetecho: https://meetings.conf.meetecho.com/ietf116/?group=cats&short=cats&item=1
Zulip https://zulip.ietf.org/#narrow/stream/cats

Note takers: Linda Dunbar, Cheng Li

#1 15:30 5mins Title: Intro, WG Status
Presenter: Chairs
No comments after Chair presentation.

#2 15:35 5mins Title: Quick review of charter deliverables and milestones
Presenter: Chairs
[Adrian]Reminding people this is a working group, not BoF. Even though there are a lot of drafts, Chairs are focusing on the deliverables first. One step is to adopt the problem statement, use cases, gap analysis, and requirements. Those drafts are not necessarily going to RFCs, they are mainly to guide the work of the WG.
The charter talks about the Groundwork. This is trying to achieve what we want to achieve.
The biggest requirement is to understand what is "Computing", how to characterize the "computing". Then it can open the door for evaluating existing tools to see if they are useable or changes needed.
We must not forget there are management tools.

#3 a 15:40 10mins Title: Problem statement and use cases
Presenter: Kehan Yao
Draft: https://datatracker.ietf.org/doc/draft-yao-cats-ps-usecases/
[Luay Jalil, Verizon]Are there any network constraints like bandwidth?
[Kehan]There is image recognization that needs high bandwidth requirement.

#3 b 15:50 10mins Title: SDWAN and SFC use cases
Presenter: Shuai Zhang
Drafts: https://datatracker.ietf.org/doc/draft-zhang-cats-computing-aware-sdwan-usecase/
https://datatracker.ietf.org/doc/draft-zhang-cats-computing-aware-sfc-usecase/
[Luay Jalil, Verizon]To make it more realistic for me, one Cloud can have multiple sites.
[Jim Guichard]SFC: there is a third component: policy (in addition to computing). Please take it into consideration.
[Joel Halpern] The SD-WAN use case fits well with CATS. But SFC is not quite aligned with CATS. CATS is more about user traffic to different service instances. The SFC is about SF within the network. It is an interesting problem, but not within the CATS scope.
[Adrian]There are three types of use case: one use case type motivates the work; one answers the question, "With the technology, what else can do:; and one is future-looking: What might we extend this to in the futue? I would urge the authors of these use case documents to come together to sort out what is in the scope and what is out of the scope.
[Dean Bogdanovic]Some use cases are realistic, some are not. E.g., the Car use case is never possible. Please focus on the use cases that are realistic.
[Adrian]Good point

#4 16:00 10mins Title: Gap Analysis and Requirements
Presenter: Daniel Huang
Draft: https://datatracker.ietf.org/doc/draft-yao-cats-gap-reqs/
[Adrian]Maybe the requirement part should be lumped together with the Problem Statements. Please think about it and discuss with the co-authors.

#5 16:10 15mins Title: Framework and Architecture
Presenter: Cheng Li
Draft: https://datatracker.ietf.org/doc/draft-ldbc-cats-framework/
[Chong Feng Xie]Concerned about the term "Computation". What is "computing", what is "computation"?
[Cheng Li]Excellent question. Could you please send the email to the list or raise an issue in github?
[Jim Reid]Are these drafts going for WG adoption?
[Adrian]The WG just started. This the meeting and the discussion is to find out what drafts we have.
[Liu Xiang: Huawei] Abuot "Path Intelligence", I am wondering if the Path Computation can only be achieved by "Path Intelligence unit", can this be deployed in routers?
[Chang Li]This is a logical component, which can be deployed in routers or by itself.
[Liu Xiang]For "Instance ID", why call it "Binding ID"? Why not call it "Instance ID"?
[Cheng Li]Good point. let's discuss on the mailing list.

#6 a 16:25 10mins Title: Compute Resource Modeling consideration
Presenter: Kehan Yao
Draft: https://datatracker.ietf.org/doc/draft-du-cats-computing-modeling-description/
[Cong Li]Computing resources is important for CATS. The first level of Hardware (CPU/GPU/etc.) are important for DC. Should we also take DPU into the consideration?
[Kehan]Your question is more to HW. That is hard to characterize, especially difficult to make it relevant to the ingress node. We will consider it.

[Dean Bogdanovic]For the computing model, there was presentation by DMTF (Distributed Management Task Force) in IETF98 that can be helpful to this work.
[Adrian]That is a really good pointer. ALTO also needs this set of metrics.

#6 b 16:35 10mins Title: Compute Metric distribution
Presenter: Hang Shi
Draft: https://datatracker.ietf.org/doc/draft-shi-cats-analysis-of-metric-distribution/
[Adrian]This draft is partly Framework draft, partly Problem Statements.
[Chuan Ma]I would like to see a complete version of the draft. So that the draft can guide the Framework draft.
[Dirk Trossen]This draft might be helpful to guide the framework. Some of the findings can be lumped into framework draft.
[Adrian]Yes.
[Dirk Trossen]Would be nice to get feedback from the community.

#7 16:45 10mins Title: Open Discussion
ALL

#8 16:55 5mins Title: Next Steps
Presenter: chairs
[Adrian]We Chairs need to talk to ALTO chairs. We will discuss with the AD whether energy use is a metric that is in scope.
[Erik Nordmark]The problem statement is about integrating Computing metrics with Network Metrics. But I don't see anywhere describing how to integrate them.
[Adrian]The WG discuss the metrics and how to collect/distribute them. We don't normally standardise software and maybe the vendors want their own method to do the path selection.