Google proposal: Nephio v1 user stories

[Kaushik Bhandankar]

Hello SIG-arch community members,

As discussed during the last SIG-architecture meeting, we are sharing Google's proposal of the Nephio v1 user stories. Please go over it and add your comments prior to the upcoming SIG-arch meeting.

Thanks!

[Eric Debeau]

Hello

I have not the rights to modify the document.

Please find some remarks

Objective

User stories should focus to identify the new functional blocks/features that are currently missing in K8S or the ecosystem to achieve the Zero Touch Automation promise.

I agree that using 5GC can be a good example to illustrate the complexity that we may face in Telco deployment, but Nephio activity should not focus on solving some networking issues. As a result, we need to get some 5GC experts.

User story 1:

What is the goal with regards to automation ?

There is a coredns demo where some features are demonstrated. Using the 5G core network function should be used only to demonstrate features not demonstrated by coredns:

- ie capability to model how various CNF can use same source of truth (eg plmnid, tac, sst...) => typically how to create a source of truth for global parameters that must be used to deploy the various CNF

- capability to model how to model a 'service' made of various CNF

User story 2:

[Use-case 2.1] Upgrade the resources (CPU / memory) of one of the existing UPFs => I do not understand what Nephio will bring from such user story. We can reuse vertical pod autoscaling from K8S for such requirement. In addition, we must be sure that the CNF are able to benefit from more CPU, memory.

This use-case may be relevant if the UPF is based on various PODs and that we need some 'logic' in addition to vertical pod autoscaling. But I believe that Free5GC UPF is only composed of one pod

FYI, 5GC UPF HPA is included in the 5GC helm chart: https://github.com/Orange-OpenSource/towards5gs-helm/blob/main/charts/free5gc/charts/free5gc-upf/templates/upf/upf-hpa.yaml

[Use-case 2.2] Setup a new edge cluster3 and deploy UPF => This is a good use-case that can demonstrate Nephio added value. Deploying a new UPF will require some configuration for other NF.

User story 3: Upgrade control-plane capacity

This is a good user story from Telco point of view, but that should not be limited to memory extension for a given CNF.

Typically, we need to upgrade control capacity when PDU sessions reach a limit, and it is not always directly linked to memory/CPU consumption.

From Nephio point of view, I believe that user story 2 and 3 are very similar => Upgrade the capacity for a set of CNF. We may have sub use-case for user plane and control plane.

User Story 4

Beyond a composition of individual user stories, this showcases the ability of Nephio to model and understand relationships and effect changes in concert. => +2.

There is also some requirements for configuration management. I would propose a user-story to change the configuration that may impact various CNF (eg add a new tac) .

Best Regards

Eric

--
You received this message because you are subscribed to the Google Groups "SIG Network Architecture" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sig-network-archit...@nephio.org.
To view this discussion on the web visit https://groups.google.com/a/nephio.org/d/msgid/sig-network-architecture/CAOAguQkVXvoA%2B%2BJkKDMCVH17CutPh2-3h9ZmgWt_r87LQE5zbg%40mail.gmail.com.

[Yuriy Platonenko]

Hello team,

I want to clarify Use case one.  Is the priority to leverage 5GC or to find a real commercial use case?

- If it's about leveraging open source 5GC - the use case fits quite nicely

- If we talk about a realistic operator scenario, I'm not sure there is a need to deploy a new 5G core. It also generates discussion if UPF/AMF/SMF are enough or need to add UDM, policy, charging and etc.

What about considering a Private network scenario here? it usually has limited scope and a limited number of Network Functions.

regards, iurii

вівторок, 6 вересня 2022 р. о 08:41:08 UTC+2 eric....@gmail.com пише:

[Tom Nolle (Public)]

My concern with the 5GC user story, and the others, is that they may not really involve that much in the way of agile lifecycle operation.  The real value of Nephio has to come from bringing cloud deployment and lifecycle operations to the telco world.  Don't we need to benefit the telco world with the cloud's capabilities to do that?  I'd also like to note that the NFV ISG was very conservative in their proofs of concept and that replicating that work doesn't move the ball much.

Tom

[Yuriy Platonenko]

fine with me.

I looking forward to meeting on Wednesday to get more clarity from Sana and the team 

вівторок, 6 вересня 2022 р. о 14:24:18 UTC+2 tno...@cimicorp.com пише:

[Kaushik Bhandankar]

Thank you for your early feedback (please keep it coming, these are useful conversations). Responding to the early comments:

1. How to scope 5GC (set of NFs, set of use-cases): That is up to us. I don't think this is an insurmountable problem. Google's proposal initially had PCF / UDM etc. but we decided to scope to AMF / SMF / UPF as that's the minimal Free5GC set. We can alter this based on the community input.

2. Realistic Telco scenario: Many Telcos are in the process of deploying & scaling 5G networks. This Google proposal addresses the user-stories of initial deployment and scaling.

3. Agile lifecycle operation: I am not sure I fully agree with this. If there are additional day2 user stories we want to incorporate, that's totally reasonable but 5GC is a key portion of Telco focus at present and we should prove the value Nephio value to such important areas of investment.

Like I had mentioned earlier, I look forward to working closely with Sana (as chair) and the Telcos in this community to help align on the use-cases. This email had shared Google's proposal to seed the conversations in the community.

Thank you!

To view this discussion on the web visit https://groups.google.com/a/nephio.org/d/msgid/sig-network-architecture/3a6bf656-5486-41cc-bf2c-f1f28cfbc430n%40nephio.org.