Re: [sfc] comments on draft-quinn-sfc-problem-statement-02

[Jamal Hadi Salim <[e-mail]>]

Jim,
Finally getting back to this (I am at the meeting as i type this); here are some
suggestions, section 3 - bullet #4:

---------
>Dataplane Metadata: Data plane metadata provides the ability to
>exchange information between the network and service functions,
>between service functions, and service functions and the network.
>Metadata can include the result of antecedent classification,
>information from external sources or forwarding related data.
>For example, service functions utilize metadata, as required, for
>localized policy decisions.

to:

>Dataplane Metadata: Data plane metadata provides the ability to
>exchange information between the network and service functions,
>between service functions, and service functions and the network.
>Metadata can include the result of antecedent classification,
>information from external sources or forwarding related data.
>For example, service functions utilize metadata, as required, for
>localized policy decisions
or global (across service chain) pipeline stage processing indices.

YMMV, I feel it may require more text as to what is meant as "pipeline
stage processing index".

The one more requirement we have may be more solution space on achieving
pipeline indexing, but could fit in the problem statement, to be specific:
we have a requirement to be able pass arbitrary metadata between
nodes (means desire to have more than just basic 32/64 bit constructs).

It may mean adding one or two more sentences above.
Thoughts?

cheers,
jamal




On Tue, Jan 14, 2014 at 8:19 AM, Jim Guichard (jguichar)
<[e-mail]> wrote:
> Hi Jamal,
>
> Thanks for this input. Comments inline.
>
> On 1/14/14 6:16 AM, "Jamal Hadi Salim" <[e-mail]> wrote:
>
>>Some context:
>>ForCES is defining what is known as the inter-FE LFB
>>(http://tools.ietf.org/html/draft-joachimpillai-forces-interfelfb-03)
>>to describe chaining of network functions.
>>
>>One of the desires ForCES has is for horizontal scaling of
>>network functions/service.
>>We would like to take a network function or service and split
>>its parts across several processing resources. This requires to
>>pass around relevant "Pipeline stage indices"
>>Brad McConnell did touch on it here (slide 6):
>>http://www.ietf.org/proceedings/87/slides/slides-87-nsc-6.pdf
>>
>>This view is missing from the problem space.
>
> Jim> I agree. Do you have suggested text that can be reviewed and
> discussed further?
>
>>
>>Additionally - we have a requirement to be able pass arbitrary
>>metadata between nodes (means desire to have more than just
>>basic 32/64 bit constructs)
>
> Jim> I think this is an area that will require some discussion as part of
> the SFC architecture. The problem statement has no "solution" text and
> simply states that metadata between the network and SF's, and between
> SF's, is needed. The format of that metadata should be discussed outside
> of the problem statement.
>
>>
>>cheers,
>>jamal
>>_______________________________________________
>>sfc mailing list
>>[e-mail]
>>https://www.ietf.org/mailman/listinfo/sfc
>