Re: [8023-CMSG] Questions
I personally don't think PAUSE should be thrown out. However, I do believe that PAUSE is rarely used because it has effects that some people unfortunatley don't expect (e.g. queues back up).
That said, PAUSE does not really break my personal paradigm becaue I don't see PAUSE as providing CoS/QoS. I view PAUSE as taking a big pipe and making it a smaller pipe. It does not introduce multiple traffic classes, or differentiated traffic handling - it simply results in decreasing the maximum bandwidth that can be supported by a link, the same as running a port in 10M mode vs 100M mode would do. QoS/CoS is still supported above the MAC, and attempts to make traffic differentiation decisions based upon configured policies that make the best use of whatever bandwidth is available in that single, albeit slower, pipe.
- Matt
> -----Original Message-----
> From: Booth, Bradley [mailto:bradley.booth@INTEL.COM]
> Sent: Sunday, May 16, 2004 11:42 PM
> To: STDS-802-3-CM@LISTSERV.IEEE.ORG
> Subject: Re: [8023-CMSG] Questions
>
>
> Matt,
>
> Thanks. I understand the structure a bit better now. Is it also your
> belief that 802.3 should throw out PAUSE? That was added for some
> QoS/CoS functionality, but it would obviously break your paradigm.
>
> Thanks,
> Brad
>
> -----Original Message-----
> From: Matt Squire [mailto:MSquire@HatterasNetworks.com]
> Sent: Sunday, May 16, 2004 10:24 PM
> To: Booth, Bradley; STDS-802-3-CM@LISTSERV.IEEE.ORG
> Subject: RE: [8023-CMSG] Questions
>
>
>
> I'll throw some opinions out, may or may not agree with Norm.
>
> >
> > Norm,
> >
> > Thanks for the response. Two follow-up questions:
> > 1) Is it understood or implied that Ethernet knows how to
> > direct frames
> > to and from these 8 queues?
>
> 802.1 knows how to direct frames to those queues. The queues
> are above
> the MAC and therefore Ethernet knows nothing about it.
>
> > 2) What if the device does not use a bridge as in an adapter?
> >
>
> A device, even a host adapter, can use 802.1 rules if it
> wishes to apply
> layer two prioritization. Alternatively, a device could use layer 3
> rules to perform prioritization.
>
> I would argue that the main point is that there are a large number of
> prioritization and scheduling options available above the MAC
> to ensure
> a wide variety of services are possible, and that all of these operate
> best when Ethernet remains a single pipe with no inherent QoS/CoS.
>
> >
> > -----Original Message-----
> > From: owner-stds-802-3-cm@LISTSERV.IEEE.ORG
> > [mailto:owner-stds-802-3-cm@LISTSERV.IEEE.ORG] On Behalf Of
> > Norman Finn
> > Sent: Sunday, May 16, 2004 11:11 AM
> > To: STDS-802-3-CM@LISTSERV.IEEE.ORG
> > Subject: Re: [8023-CMSG] Questions
> >
> >
> > Brad,
> >
> > I think you did miss the mark, particularly with:
> >
> > "Considering that Ethernet doesn't know in advance about the
> > provisioning
> > of the network and does not care about which packets it delays or
> > drops,
> > then it is likely that 802.1 and the upper layers can do all the
> > priorities or differentiated services that they want but will see
> > diminishing returns as the load on the network increases."
> >
> > I would agree with, "Ethernet doesn't know in advance about the
> > provisioning
> > of the network", but 802.1D bridges certainly do care about
> > which frames
> > are
> > delayed or dropped. Bridges define the use of 8 queues per
> > output port,
> > and
> > frames are marked with 8 levels of priority. Although
> strict priority
> > scheduling is the only queue draining algorithm specified in the
> > standard,
> > others are explicitly allowed, and most vendors implement
> > varieties that
> > provide very good latency and bandwidth guarantees. Furthermore, a
> > great
> > many bridges are able to assign priorities to 802.3 frames based on
> > criteria
> > such as IP DSCP code points.
> >
> > In short, ethernet is *far* from "best effort".
> >
> > -- Norm
> >
> >
>