Minutes IETF125: cats: Mon 06:00
minutes-125-cats-202603160600-00

Computing-Aware Traffic Steering (cats) WG Agenda - IETF 125
When: 2026-03-16 14:00-16:00 Shenzhen
Co-Chairs: Adrian Farrel, Peng Liu
Secretary: Cheng Li
Logistic
Materials: https://datatracker.ietf.org/meeting/125/session/cats
Note taking: https://notes.ietf.org/notes-ietf-125-cats
Meetecho:
https://meetings.conf.meetecho.com/ietf125/?group=cats&short=cats&item=1

Zulip: https://zulip.ietf.org/#narrow/stream/cats

Attendees number: 70

1. Agenda Bashing, Introduction, & WG Status (5 min)
   Chairs

[Peng] Agenda bashing.
1.The meeting will focus on the WG documents, including CATS framework
and metric definition drafts.
2.Also, new topics such as about implementing CATS in Optical network
can be discussed as well.
Use cases and requirements document has passed the IESG Last call,
awaiting for CATS framework draft(depending on this draft) to publish as
RFCs together.

1. WG Documents Update
   2.1. CATS Framework Update (10 min)
   Cheng Li
   https://datatracker.ietf.org/doc/draft-ietf-cats-framework/

[Adrian]Thanks Cheng for hosting social event!
[Adrian]The draft will be sent to the IESG Telechat on 2nd April, hope
we can move the draft smoothly.

2.2. CATS Metrics (10 min)
Kehan Yao
https://datatracker.ietf.org/doc/draft-ietf-cats-metric-definition/
No input on question on the slides.
[Peng] We need to move this to the list
[Peng] If no more comments we will talk to the AD about whether we
take this forward for publication as an RFC

1. Upcoming Documents
   3.1. CATS OAM Framework (10 min)
   https://datatracker.ietf.org/doc/draft-fu-cats-oam-fw/
   Quan Xiong

[Haomian Zheng] The OAM framework is very useful. We have some
considerations on the layer of OAM. Because in the metric document, it
defines L0, L1, and L2 metric, not layer. I would suggest to use
consistent terms across documents. Moreover we will also evaluate the
necessity to introduce other network layer OAM technologies.

[Adrian] Important to be working on OAM. Thanks.
We need to distibguish between OAM for the "CATS system" and OAM for the
network that supports CATS. We should not need to work on the network
OAM, but build on tools that already exist.

3.2 CATS Security Considerations (10 min)
https://datatracker.ietf.org/doc/draft-wang-cats-security-considerations/

Jinyu Shi

[Jim Guichard] Please look into the security consideration section in
the CATS framework, to sync up the work.
[Peng] Please prioritize the work, because you have listed so many
security considerations. It will be better to prioritize the work.
[Jinyu] Sure.

3.3. Yang Model for CATS (10 min)
https://datatracker.ietf.org/doc/draft-yl-cats-data-model/
Quan Xiong
[Adrian] The WG currently has 3 WG documents, and 2 of them are close
to RFC, and one close to WGLC. The chairs will discuss what is the next
topic the WG should focus on, and consult the WG.

3.4. Computing metrics as a service (CMAS) for facilitating traffic
steering in CATS framework (10 min)
Yina Dai
https://datatracker.ietf.org/doc/draft-zhangb-cats-cmas/

[Guanming Zeng] The draft separates computing metric advertisement
(via CMAS) from network topology. However, optimal path selection
requires a consistent view of both compute and network states. How do
you handle the synchronization between CMAS updates and routing protocol
convergence?

[Yina] For The update frequency from the CMAS, The framework only
consider the usage of the slot. When the service site has its metric
update, it just give us the GAS change, we will change our slot.
[Yina's teacher] SDN controller has the TE-DB to know the the topology
and link states. Service metric table is a light-weight table, will not
bring burden to the CPS. We have implemented the system, it is important
to run the demo to verify the design.

[Quan] What's the difference between L1 and L2 metrics in this model?

[Yina] L1 is the normalized physical metric. Our metrics do not care
about L1. The network does not need to know the details.
[Peng] Please move the discussion to the list. From WG aspect, we may
need to find out what can we do in the CATS WG. You might need to find
out what you want to do in this WG. We did not see what do you want to
contribute in the CATS now.

1. Disscussion of Protocols Applicability
   4.1. Protocols Applicability for Computing-Aware Traffic Steering
   (CATS) (10 min)
   https://datatracker.ietf.org/doc/draft-yxl-cats-protocols-applicability/

   Huijuan Yao

[Xinxin] It is a useful work. Will the draft recommends the BGP-LS as
a recommended solution?
[Huijuan] This is an individual draft, and only the first revision.
This draft will not recommend solution, but only analyzing the protocols
if it can be applicable to CATS.

[Jeffery] BGP-LS might not be a good choice for distributing fast
changing data.

[Ketan from the chat] Thanks for taking up this aspect. It would be
very helpful to do this analysis as it will assist possible next
steps/work that needs to be evaluated by the protocol-specific WGs.

[Jim] Which is why protocol specific work is premature at this point.

[Adrian] Now is not the timing for picking the protocol solution, but
it can trigger participants to think and understand volume and churn.
[Jeffery] BGP-FS might not be a good choice as well, because it is not
fast enough.

4.2. BGP Edge Metadata Path Applicability (10 min)
https://datatracker.ietf.org/doc/draft-dunbar-cats-5g-metadata-applicability/

Linda Dunbar
[Guanming] 5G radio conditions and UE mobility states can change very
rapidly (sub-second), whereas the CATS control loop involving metric
collection, path computation typically operates on a slower timescale.
What mechanisms do you suggest to handle this 'timing mismatch'?
[Linda] Only present a work in IDR. I have nothing to suggest to
handle this problem. This problem may be beyond the scope of this draft
and be discussed by other drafts.

1. CATS and Optical
   5.1. Framework and Applicability of CATS in Optical Transport
   Networks (OTN) (10 min)
   https://datatracker.ietf.org/doc/draft-zhao-cats-otn-applicability/

   Xiao Li
   [Adrian] It is about applicability of OTN to CATS. Should think
   about how to build OTN for CATS. Should consider to integrate ACTN
   and other work. We may need to pay attention to work, and might let
   CCAMP people to get involved with this work. Good start.

5.2. Unified Optical Networks and AI Computing Orchestration (UONACO)
(10 min)
Problem Statement, Use Cases and Requirements
(https://datatracker.ietf.org/doc/draft-tan-ccamp-uonaco-problem-statement/)

Control Framework
(https://datatracker.ietf.org/doc/draft-hu-ccamp-uonaco-control-framework/)

Zheng Han
[Adrian] A reminder that this would be some use cases for Layer 1 VPN?

[Haomian] This is some kind of Layer 1 VPN connecting CATS ends.
[Adrian] Will drop an email to the list about this topic. You can read
some Layer 1 VPN related RFCs, it might help.

1. If there is time
   6.1. Intelligence Delivery Network (IDN) (5 mins)
   Framework of IDN for Deep Learning Inference
   (https://datatracker.ietf.org/doc/draft-wang-cats-idn/)
   Framework for Large Language Model Inference
   (https://datatracker.ietf.org/doc/draft-wang-cats-odsi/)
   Hanling Wang

[Chuanyang] CDN has a core function-loading balance. But CDN has the
loading balance implemented in application layer. For IDN, how to
implement loading balance? Application layer solution like CDN? Or use
CATS as the network layer solution? Which is better?
[Hanling] Intent to use the application layer loading balance like
CDN. But for IDN, it needs to consider how to decompose LLM into small
parts regarding to the different aspects. Continue discussing it
offline.

6.2. In-Network Scheduling & Semantic Contract for AI Workloads (5 mins)

IntelliNode: In-Network Intelligent Scheduling Extensions for
CATS(https://datatracker.ietf.org/doc/draft-li-cats-intellinode-network-scheduling/)

Semantic-Driven Traffic Shaping Contract for AI
Networks(https://datatracker.ietf.org/doc/draft-li-cats-aisemantic-contract/)

Teng Gao

[Peng] Please find out the correct WG to move your draft. Or find the
relation between the work and CATS WG.

1. Open Disccusion (5 mins)
   [Adrian] We need to stick to the Charter, please work within the
   chart, it will help us to finish the work more efficient.
   [Kireeti] Was the last presentation in the charter?
   [Adrian] We need to hear from people of what people are doing
   right now if we have time. It does not require to be in the charter.
   Right now, the work is not in the Charter.
   [Adrian] See you in the next meeting.