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

Re: [EFM] Half-duplex deferral for MAC-PHY rate matching a


Arthur has already proposed a scheme using deferral which will work with all current
MACs. As has been pointed out the limitation of Arthur's scheme is that it is
genuinely half-duplex and so may only support an aggregate data rate of 100Mbps.
However, Shimon Muller pointed out that any MAC which is designed according to the
specification will work with deferral (as Arthur suggests) but may also receive
while transmitting in half-duplex mode. Some of us think there is some risk in
assuming that all existing MACs will work like this and so we would like to make
this operation optional in the PHY. However, we can make it obligatory that future
MAC designs work in this manner by making the MAC specification more explicit.

Let me summarize:

1. We have a proposal that we know will work for every existing MAC.

2. We have an augmentation that will work for most existing and all future MACs
which allows 100Mbps full duplex data transfer. That is as fast as you will get with
an MII interface. It's perfect - you cannot do any better.

Remind me again, why do you want to invent more schemes?


christophe.del-toso@xxxxxx wrote:

>      Vladimir, Steven, Arthur and all,
> I think that needs for more than 100 Mbps already exist today (at least in some
> regions of the world) and that generally speaking loop aggregation is a smart an
> appealling concept which can get a lot of momentum.
> Moreover,  deploying fiber to offer 100 Mbps on short loops while in some
> deployment scenario VDSL PHYs can do the job is not a cost effective solution.
> It is proven by theory and confirmed by early field trials (In ST, we did
> several field tests with a VDSL lab-prototype) that on short loops VDSL can
> offer aggregate bit-rates above 100 Mbps.
> I think it would be a pity to limit ourselves at the interfaces with higher
> protocol layers when we have today a PHY that can offer this possibility.
> I think we could regret it in the long run even if I should recognize that this
> business case will not be the most deployed but it will exist.
> I think that the MAC-PHY rate adapation through CRS is a smart solution for our
> problem in the short-terms. It will then allow VDSL PHYs to interface directly
> with conventional and standard ethernet MACs.
> In the other hand, I am wondering myself if, by anticipating long-term
> requirements, the group should also investigate the possibility to extend
> capabilities of MII interfaces by adding few signals.
> On this point I remember a presentation from Michael Beck (Alcatel) (in the
> Austin meeting) presenting the concept of VMII.
> I think that the first objective could be MAC-PHY rate adapation through CRS and
> that second objective could be something of the kind of VMII or something less
> but something that lets maximum freedom.
> Your comments are welcome !
> Regards
> Christophe Del-Toso, ST Micro
> ______________________________ Reply Separator _________________________________
> Subject: Re: [EFM] Half-duplex deferral for MAC-PHY rate matching a
> Author:  oksman (oksman@xxxxxxxxxxxx) at internet
> Date:    12/12/2001 01:23
> Arthur,
>       I would support Steven's idea. For a full duplex 100 Mb/s customer fiber w
> ill probably work better. I would like to notice that the expected aggregate cap
> acity of a VDSL link with loop length zero is close to 100 Mb/s. The first hundr
> eds feet will already reduce it significantly.
>       If we really consider a business case with a distance of 200-300 ft, I wou
> ld look for a fiber solution.
> Vladimir.
> Steven.Haas@xxxxxxxxxxxx wrote:
> > Arthur,
> >
> > Our experience has shown that the half duplex method is the safest way and wor
> ks with most MACs. The CRS can always be used to control the flow of frames from
>  the MAC to the PHY, even at all the different rates proposed for EFM copper obj
> ectives.
> >
> > The only drawback is the "limitation" of 100Mbps. This can be overcome with mo
> st MACs using a flow control message (802.3x). This method is not watertight and
>  not all MACs process this in a timely fashion so buffers are needed in the PHY.
>  The case of an EFM PHY working above 100Mbps will most likely only happen in ve
> ry short range loop aggregation scenarios when there are multiple lines going to
>  a business customer. I expect fiber to be used in these cases so creating a new
>  MAC seems unnecessary.
> >
> > I propose the following engineering tradeoff:
> > 1. Use of the half duplex method for most modes of operation. This covers all
> rates up to 100Mbps aggregate.
> > 2. Usage of flow control, combined with the appropriate buffers, in the rare c
> ases of rates above 100Mbps.
> >
> > A third option is to create a jump in supported service rates from 100Mbps hal
> f duplex to 100Mbps full duplex.
> >
> > Steven
> >
> > -----Original Message-----
> > From: Rich Seifert [mailto:rich@xxxxxxxxxxxxxxx]
> > Sent: Monday, December 10, 2001 7:25 PM
> > To: Behrooz Rezvani; 'Arthur Marris'; 'Shimon Muller';
> > Subject: RE: [EFM] Half-duplex deferral for MAC-PHY rate matching and comp ati
> bility with existing silicon
> >
> > At 7:17 AM -0800 12/10/01, Behrooz Rezvani wrote:
> > >Arthur,
> > >
> > >As I mentioned in our conference call we need to support data rate greater
> > >than 100 Mbps in total. As I have been listening to you, Rich and Shimon, I
> > >gather that there is a way to operate the MAC such that we can exceed the
> > >half duplex rate.
> >
> > Operating at data rates in excess of 50 Mb/s (full duplex) would be
> > problematic with the system being proposed so far. Since it is
> > predicated on a single, 100 Mb/s MAC operating in half-duplex mode,
> > the combined transmit+receive rate cannot exceed 100 Mb/s; this is
> > the equivalent of a 50 Mb/s symmetrical full-duplex PHY.
> >
> > In order to operate at greater data rates, one would need to use a
> > Gigabit MAC operating in half-duplex mode. Such MACs are relatively
> > rare; indeed, even if they exist, the use of half-duplex GbE is more
> > theoretical than practical--there are no GbE repeaters in commercial
> > use. It is not even clear that such MACs work properly in half-duplex
> > mode.
> >
> > In addition, it is not possible to aggregate multiple 100 Mb/s MACs
> > when operating in half-duplex mode. The current Link Aggregation
> > standard restricts aggregation to full-duplex links only.
> >
> > >Note that VDSL PHY is full duplex system, and it can transmit and receive
> > >independently.
> > >
> >
> > I suspect that EFM will want to operate over a variety of PHY types
> > and speeds. Rather than trying to cobble all of these systems to some
> > pre-existing MAC chips (which I agree may provide some short-term
> > benefit), perhaps it would be wiser in the long run to define a
> > full-duplex MAC with a variable (quasi-static) data rate.
> > --
> >
> > --
> > Rich Seifert                    Networks and Communications Consulting
> > rich@xxxxxxxxxxxxxxx            21885 Bear Creek Way
> > (408) 395-5700                  Los Gatos, CA 95033
> > (408) 395-1966 FAX