MAM Definition

["Francois Le Faucheur (flefauch)" <flefauch@cisco.com>]

Hello,

The current MAM definition purely specifies one Bandwidth Constraint per
CT (as per -diff-te-reqts and diff-te-mam). It was proposed that this
definition be enhanced by applying one additional bandwidth constraint
which would be an aggregate constraint across all CTs (in fact it
appeared that a few people had been assuming this was already part of
the definition).

Just to illustrate this, with the new definition one could ensure, on a
given link, that:
	Reserved (CT0=Voice)              <=   CT0=50%-of-link
	Reserved (CT1=Premium)            <=   CT1=40%-of-link
	Reserved (CT2=Best-Effort)        <=   CT2=60%-of-link
      Reserved (CT0+CT1+CT2=everything) <=   Max Reservable=120%-of-link

I don't have objections to adding this enhancement to the MAM
definition.

Other opinions?

Assuming we go down that path, one issue we would need to address is the
fact that the diff-te-reqts draft is going to be published as an RFC and
will mention MAM with its "old" definition, which would result in
inconsistency and confusion. One simple solution to this would be to
give a different name to the model with the "new" definition. I don't
have a specific proposal, but I'm sure we could come up with some name
which better reflects the fact that we have both a per-CT constraint and
an aggregate constraint (better than "Maximum Allocation"). Come to
think of it, I do have a proposal, maybe something like "Separate and
Aggregate constraints Model (SAM)".

Thoughts?

Cheers

Francois

PS: Changing MAM definition actually has quite a few ramifications in
the -proto draft (eg about advertisement of Max Reservable and BC
sub-TLV) and probably some in -russian draft. I will send a separate
email with a detailed proposal on how to reflect this.