Re: [sfc] Progression of use case documents in the SFC WG

Dear all,

1+ ---- I also fully support the WG director's approach for domain specific use case drafts.

In my humble opinion Jim and Thomas proposed the most pragmatic and most efficient way to work out, how network and DC operators utilize SFCs today. To fully understand, we have to analyze and document the domain specific characteristics of today's more advanced SFC deployments, their potential weaknesses and what are the expectations on future-oriented approaches the WG is working on. A detailed understanding of the real thing is key to successful work.

Even though e.g. DCs and mobile networks may be part of one and the same service delivery chain, their SFC use cases and requirements may be quite different. Even if we claim they both use the same middleware boxes/middleware VMs. On the other hand, typical operator use cases for fixed networks are (very often) a small subset of what is seen in mobile. The difference between them is basically in the amount and representation of metadata.

If we have a more clear picture on the domain specific use cases including their metadata dependence and metadata handling we may proceed to try to generalize the documented use cases to support a flexible enough,  but still unique and future proof  SFC specification. I already had a f2f meeting with Hongyu Li last week and suggested such an approach.

Best regards,
Walter

Von: sfc [mailto:sfc-bounces@ietf.org] Im Auftrag von Thomas Nadeau
Gesendet: Freitag, 28. März 2014 13:53
An: Ken Gray (kegray)
Cc: Guichard Jim; mohamed.boucadair@orange.com; sfc@ietf.org
Betreff: Re: [sfc] Progression of use case documents in the SFC WG

            I agree with the chairs' proposal as well.  Lets go with a small number and focus things down here a bit.

            --Tom

On Mar 27, 2014:8:28 AM, at 8:28 AM, Ken Gray (kegray) <kegray@cisco.com<mailto:kegray@cisco.com>> wrote:

I'd give a +1 to the chairs ...

If there is going to be more than one document (and we seemed hell bent on more than one at the BoF ... we could, as a group, settle on "a small number"), that they have some focus.

I don't find the focus or organization described by the chairs onerous - in fact, GIVEN that we have the liaison(s) in place and that they do want a voice here, and that at least one of them has a "domain focus" I find it logical to start with broadband and develop use cases in a set of non-trivial domains.

They had to define "a small number" ...or "a small number" becomes a big number.  Now we know how many "a small number" is.

If draft-liu is stripped of enough content by the categorization of the "small number" it's efficacy should be questioned.  To your specific point, as a group we can decide on moving the specific text you mention back to Problem or otherwise re-home it.  It shouldn't be the sole reason draft-liu exists.

I would have gone a bit further than the chairs, frankly.

There is SO much use case literature out there right now, I frankly don't want to see the IETF repeat any more than it has to.  I would hazard that most of us have read these things before in one of several forums.

So, IMO, the use cases should provide support for the problem statement and the development of a proposed header functionality (solution), and as such should illustrate significantly unique requirements ...so that we can assess the efficacy of the proposed solutions.  I hope the "owner by area" described by the chairs will take on the responsibility of making sure that their examples are significantly unique for  consideration and addressable in the solution.

Because their work represents the work of many (whole organizations) and is hopefully further distilled by the author here into unique examples, in fairness their contributions should probably be credited to "various" with a nod to their organization.

And, because these use cases have been reviewed in other forums by many people, the process should be more expeditious.

From: "mohamed.boucadair@orange.com<mailto:mohamed.boucadair@orange.com>" <mohamed.boucadair@orange.com<mailto:mohamed.boucadair@orange.com>>
Date: Thursday, March 27, 2014 3:07 AM
To: "Jim Guichard (jguichar)" <jguichar@cisco.com<mailto:jguichar@cisco.com>>, "sfc@ietf.org<mailto:sfc@ietf.org>" <sfc@ietf.org<mailto:sfc@ietf.org>>
Subject: Re: [sfc] Progression of use case documents in the SFC WG

Dear chairs,

Some comments below:
·         The proposed actions are not aligned with the feedback received in this thread (Progression of use case documents within the SFC working group). Answers to that poll are in favor of having a generic use case document. IMHo, it is not fair to ignore what was voiced for by wg members in the mailing list as part of a formal call with clear questions.
·         Some of the text that was adopted by the WG as part of the Problem Statement (use case as part of the Problem statement) has been moved to the generic use case. That text is governed by this charter text: "1. Problem Statement: This document will provide a summary of the
problem space to be addressed by the SFC working group including
example high-level use cases. Additionally, the working group will
normalize nomenclature and definitions for service function chaining.". What to do for that text?
·         Having the generic use case document and some few detailed ones do not conflict. It is only a matter of scoping.

Given what is stated above, I disagree with your proposal.

If I have to choose (again), I would vote for having one single use cases document. Having one single document will help focusing on core aspects and would simplify the wg activity: review, last calls, etc.

Cheers,
Med

De : sfc [mailto:sfc-bounces@ietf.org] De la part de Jim Guichard (jguichar)
Envoyé : mercredi 26 mars 2014 18:54
À : sfc@ietf.org<mailto:sfc@ietf.org>
Objet : [sfc] Progression of use case documents in the SFC WG

WG:

In a message back in January, we (the chairs) proposed that the SFC WG handle the topic of use case documents as follows:

1) Have the WG develop one use case document that documents a small number of representative use cases.  The document presented by Hongyu Li at the Vancouver BOF could serve for this purpose
(http://datatracker.ietf.org/doc/draft-liu-service-chaining-use-cases).

3) For additional use cases not covered in 1) above, allow for a small number of documents that are applicable to specific
environments (e.g.  mobility, data center, broadband, and so forth.) These documents would provide more detailed information and applicability of SFC to these specific environments, and would need to go beyond what is covered in the general use case document (1). Note that it is not the intention to have every potential use case documented.

Since then, and based on the presentations/discussion in London, it appears that we have a number of documents that warrant being developed as standalone documents. Specifically:

1) A use case document on mobility, e.g., http://datatracker.ietf.org/doc/draft-haeffner-sfc-use-case-mobility/

2) A use case document on Data Centers, e.g., http://datatracker.ietf.org/doc/draft-kumar-sfc-dc-use-cases/

3) Possibly a use case document on Broadband scenarios. However, use cases from a broadband perspective are being developed in the BBF (see the liaison statement at https://datatracker.ietf.org/liaison/1304/). We also have http://datatracker.ietf.org/doc/draft-meng-sfc-broadband-usecases/.  It does not seem appropriate to adopt a WG document on the topic of broadband (at least at this time) without clarifying the relationship between draft-meng-sfc-broadband-usecases and the BBF work. In addition, we would need to understand why two efforts - one in BBF and one in the IETF -- on the same topic would be appropriate. Hence, at the present time, we do not intend to adopt a WG document on broadband scenarios, and expect to receive primary guidance on this topic from the BBF.

That leaves: http://datatracker.ietf.org/doc/draft-liu-sfc-use-cases/, a more general document. But that document includes text on three topics that would be covered in more detail elsewhere (broadband, mobile, and DC). While this document could contain pointers to the other documents, that leaves the document with very little standalone content -- raising the question of what should be done with it, or what content it could incorporate in order to be worthwhile as a standalone document.

Thus, the chairs recommendation at this time is:

1) Call for WG adoption of draft-haeffner-sfc-use-case-mobility-00.txt and draft-kumar-sfc-dc-use-cases-00.txt as WG documents (target: informational).

2) Defer action on draft-liu-service-chaining-use-cases<http://datatracker.ietf.org/doc/draft-liu-service-chaining-use-cases>  and draft-meng-sfc-broadband-usecases<http://datatracker.ietf.org/doc/draft-meng-sfc-broadband-usecases/> per the above discussion.

Does this make sense?

Jim & Thomas
_______________________________________________
sfc mailing list
sfc@ietf.org<mailto:sfc@ietf.org>
https://www.ietf.org/mailman/listinfo/sfc