Hi Linda,

From my perspective, the Attachment Circuit (AC) model can support ​​the 
dynamic establishment of logical connections​​ between the Edge Cloud DC 
Gateway and Provider Edge (PE), as outlined in your example. However, it ​​is 
not designed to support UCMP (Unequal Cost Multipath) load-balancing 
policies​​. The current capability of the AC model lies in ​​applying 
application-specific QoS policies​​ where required. For detailed answers, 
please see  inline responses below.

Regards,
Bo

From: Linda Dunbar <linda.dun...@futurewei.com>
Sent: Wednesday, April 9, 2025 9:57 AM
To: neo...@ietf.org; Chongfeng Xie <xie...@chinatelecom.cn>; opsawg 
<opsawg@ietf.org>
Subject: [neotec] Questions on Applying the Attachment Circuit YANG Model to 
Neotec’s Dynamic Network Scheduling Use Case

Bo and Co-authors of the AC:

I have some questions regarding how the Attachment Circuit YANG model  
(https://datatracker.ietf.org/doc/draft-ietf-opsawg-teas-attachment-circuit/ ) 
could be used to support a simplified version of the use case described in 
China telecom’s Neotec side meeting (
https://github.com/xiechf974/Neotec-Side-Meeting-IETF-122/blob/main/5-Cloud-aware%20Network%20Operation%20for%20AI%20Services-Qiong%20Sun.pdf)

Scenario:
Let’s assume there are 10 Edge Cloud sites.  An AI model for city surveillance 
(e.g., detecting traffic congestion or garbage classification) needs to be 
deployed dynamically to some of these sites in response to real-time events.

Step 1: The Cloud Manager needs to query the network connectivity 
characteristics (bandwidth, latency, topology constraints, etc.) between street 
cameras (or gateways, eNB that connect to those street cameras) and candidate 
Edge Cloud sites in order to determine the optimal locations for AI model 
deployment.

Step 2: Based on the information gathered, the Cloud Manager decides to deploy 
the City Surveillance AI module in 4 of the 10 Edge Cloud sites.


Question 1: If the selected 4 Edge cloud sites already have connectivity set up 
to their respective PEs (e.g., ACs are already provisioned), does this mean 
that only the Edge Cloud gateways need to be updated to advertise the new 
service (e.g., AI module prefixes)?
If an IGP is already running within the Edge Cloud data center, is there 
anything else the network controller needs to configure?
[Bo Wu] In general, yes. If assuming the mentioned network controller is a WAN 
controller, and
the WAN connectivity for these new services (e.g., L2/L3 VPNs and ACs) has 
already been provisioned, then the Edge Cloud DC Gateway could leverage BGP to 
advertise ​​IP prefixes for the new services​​.

Question 2: Suppose the AI modules in the 4 Edge Cloud sites need to exchange 
large volumes of data with strict performance constraints (e.g., XX Gbps 
bandwidth and YY ms latency). The goal is to have the network controller 
dynamically adjust UCMP (Unequal Cost Multipath) load-balancing algorithms on 
all the nodes along the paths interconnecting those 4 sites.
Can the Attachment Circuit YANG model be used to achieve this dynamic path 
optimization?
(I think that AC YANG model is NOT meant for this purpose, am I correct?)  Does 
IETF have specified those YANG model to achieve this goal?
[Bo Wu] I think the AC model is not suitable for dynamic load balancing 
policies. Furthermore, I'm uncertain whether cloud managers would request such 
load balancing policies to WAN controller. Are load balancing policies 
considered as implementation-specific techniques of WAN controllers? Or, might 
cloud managers instead require specific bandwidth, latency, and packet loss 
guarantees for individual flows (e.g., 5-tuple-based flows), rather than 
requesting load balancing policies directly?

Would be great to hear your thoughts on how these models might evolve to 
support this use case, or whether augmentations to the AC model are being 
considered in this direction.

Thank you very much

Linda

From: Wubo (lana) 
<lana.wubo=40huawei....@dmarc.ietf.org<mailto:lana.wubo=40huawei....@dmarc.ietf.org>>
Sent: Thursday, March 20, 2025 12:21 AM
To: neo...@ietf.org<mailto:neo...@ietf.org>
Subject: [neotec] Some background information on cloud-network interfaces

Hello everyone,

Here are some IETF drafts related to cloud-network interaction for your 
reference:


1)       The orchestration architecture between cloud DC and WAN networks, for 
example, the reference architecture provided by 5G IP/MPLS for 5G slicing 
scenarios.
https://datatracker.ietf.org/doc/html/draft-ietf-teas-5g-ns-ip-mpls-16#section-3.4.2


2)       Dynamic network attachments for NFs (Network Functions), for example, 
in the AC service model, how NFs Deployment, and NF Failure and Scale-Out:
https://datatracker.ietf.org/doc/html/draft-ietf-opsawg-teas-attachment-circuit-20#name-connectivity-of-cloud-netwo


3)       In current realization, APIs are consumed by Cloud systems rather than 
IETF-defined NETCONF/RESTCONF/YANG, example could be:

https://datatracker.ietf.org/doc/html/draft-ietf-grow-peering-api-00



Thanks,
Bo

发件人: Wubo (lana)
发送时间: 2025年3月19日 14:36
收件人: 'neo...@ietf.org' <neo...@ietf.org<mailto:neo...@ietf.org>>
主题: Slides of gap analysis on Telcom cloud network interfaces

Hello everyone,

Here is the Telcom cloud network interface gap analysis that I mentioned in the 
side meeting.
The link: 
https://github.com/xiechf974/Neotec-Side-Meeting-IETF-122/blob/main/6-Neotec%20gap%20analysis%20Bo%20Wu.pdf

This analysis is intended to identify the gaps between the interfaces expected 
by Neotec and the existing IETF service models and network models.
The analysis references ETSI NFV standards and documents related to WIM (WAN 
Infrastructure Manager). I think WIM interface is related to the Interface 2 we 
discussed, which covers the interface requirements and candidate IETF YANG 
models exposed from network to the cloud. I have also added some newly defined 
IETF network slices and AC service & network models to show some new candidate 
IETF YANG models.

I hope this analysis is helpful.

Thanks,
Bo
_______________________________________________
OPSAWG mailing list -- opsawg@ietf.org
To unsubscribe send an email to opsawg-le...@ietf.org

Reply via email to