Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [EFM] EFM Requirements




Carlos,

> -----Original Message-----
> From: owner-stds-802-3-efm@majordomo.ieee.org
> [mailto:owner-stds-802-3-efm@majordomo.ieee.org]On Behalf Of Carlos
Ribeiro
> Sent: Thursday, August 23, 2001 4:44 AM
> To: mattsquire@xxxxxxx; Harry Hvostov
> Cc: Lough, Andy; Stds-802-3-Efm (E-mail); carlosal@xxxxxxxxxxxxxxxxxx
> Subject: Re: [EFM] EFM Requirements

...
> I'm not aware of any widely
> deployed, complete, standards-based, solution for prioritization on the
> Ethernet level. I'm happy to be proved wrong, though, as it would be
> something really nice to have.

[Andrew] Please provide some justification for your desire to have
"Ethernet-layer" prioritisation (or other forms of QoS at this layer). But I
really wonder why you think EFM has requirements in this area that are not
general 802.3 or 802.anything requirements.

> Now, if EFM will have the mechanisms to tag packets as important
> (non-discardable, low latency, and so on), then you're right. You are
> simply mapping IP QoS over Ethernet. But even then, the standard group has
> some work to do, to make sure that the prioritization mapping is done the
> right way.

[Andrew] Note that most 802.1 switches on the market today have the ability
to look at IP-layer QoS tags, without any perceptible price increase (oops,
I didn't really mean to use that word Howard), so I do not believe there is
really much of a requirement to standardise any such mappings. IETF has
certainly looked at this mapping issue over the years with somewhat
inconclusive results (the rule tends to be that they standardise a L3->L2
mapping whenever a suitably large customer pushes for it). I would suggest
we let IETF look at this, if we decide it is needed, rather than breaking
the traditional view that IEEE 802 standards be network-layer-independent.

Matt Squire wrote:
> >The point is that the Ethernet layer has no additional work to do to
> >carry voice.  Therefore the requirements for transporting voice need not
> >be singled out - there are none.

> As I said, this is not granted - at least as far as I know. You have
> pointed out that there are several ways to carry voice; in my opinion,
some
> of them will need specific provisions in the standard. If we can map
> everything to standard features of the Ethernet, then you're right, and we
> have nothing to worry.

[Andrew] I would concur largely with Matt. Again, Carlos, you need to to
elaborate on what "specific provisions" you think are required - I assume
you are talking about non-IP-based ways. Is this perhaps a "shared vs.
switched" issue (shared media, you think, requires additional mechanisms?)?
Hasn't this debate already been had several times before in IEEE 802 with
the conclusion that it is almost always more efficient to turn a shared
medium into a switched one than it is to add complex scheduling mechanisms
to the shared medium MAC layer (see http://www.ietf.org/rfc/rfc2815.txt and
http://www.ietf.org/rfc/rfc2816.txt for some related discussion of this
topic).

> Carlos Ribeiro
> CTBC Telecom

Andrew Smith