RE: I/F locations
Paul, Rich
It does seem that the standard achieves the mythical UNIPHY for the Serial
PMD. Achieving that for the WWDM PMD is possible but something has to give.
My two cents on the matter are that both approaches (XAUI for the WAN W4,
SUPI for the LAN X4) are technically feasible and not that difficult to
define. What has to give is:
XAUI for WAN: the likely implementation of a 64b/66b PCS/WIS chip for the
serial PMD would not be usable for the WWDM WAN PHY ... this would require
vendors to supply another version of the PSC/WIS chip with the XAUI
interface to the PMD.
SUPI for LAN: the LAN X4 would have to bear the additional cost of requiring
the PCS/WIS chip (modified over the existing standard to support the
insertion of A1/A2's into the 10Gb/s data stream and the resultant rate
increase).
Not clear to me which approach wins. I'd characterize the SUPI for LAN
solution as offering the greatest opportunities for re-use and economies of
scale while the XAUI for WAN preserves what appears to be the simplest PHY
of the lot ... the currently defined LAN X4 in which the the DTE XGS can
attach directly to the PMD.
Excuse my naivity but, in immersing myself in this stuff, the question did
occur to me, what use do any of the WAN PHYs provide if they can't
interoperate with the existing SONET infrastructure? I came up with
1)existing SONET software and knowhow could be adapted to control and
monitor a 10G-W network, 2) existing OC-192 transceiver modules can now also
support 10G-W (this argument doesn't fly in the WWDM case) and 3) even
though the standard doesn't address interoperability, it will enable the
market to create it. Other than that, it seems analagous to putting one's
mail in a pre-paid Express Mail envelope and then putting that in a pre-paid
FedEx envelope ... you can do it but ...
Ted Speers
Strategic Marketing
Actel
> -----Original Message-----
> From: Rich Taborek [SMTP:rtaborek@xxxxxxxxxxxxx]
> Sent: Tuesday, December 05, 2000 1:34 AM
> To: HSSG
> Subject: Re: I/F locations
>
>
> Paul,
>
> Please see my comments below:
>
> Paul Bottorff wrote:
> >
> > Rich:
> >
> > I didn't intend to imply that XSBI was required. It is an optional
> > interface for 10GBASE-LW4.
>
> OK. To be sure, I'll make sure that I submit TF ballot comments on
> Clause 53 which reinforce this point.
>
> > Rich, I believe some or us would like to build systems using XSBI
> > interfaces. It is inconvenient
> > that 10GBASE-LR4 is unsupported using XSBI.
>
> Please feel free to use the XSBI anywhere you like. It is an optional
> interface. This means that it is not required for any P802.3ae compliant
> system.
>
> > Rich, the current SUPI implementation does not support LAN mode in the
> > UNIPHY, however this
> > could easily change by adding a periodic A1/A2 sequence to the 64b/66b
> data
> > stream. It would
> > only be necessary to add 8 A1 octets and 8 A2 octets. The current clause
> 53
> > operation can then
> > operate without modification. This would be nice since it makes the
> UniPHY
> > really unified around
> > the provided electrical interfaces.
>
> Paul, What is a UNIPHY? We discussed this in early HSSG presentations,
> but there is no occurrence of "UNIPHY" in P802.3ae D2.0 (I checked).
>
> SUPI only supports a data rate of 9.952 Gbps. A P802.3ae LAN requires a
> 10 Gbps data rate and a 10.3125 Gbps line rate when using 64b/66b
> coding. What line rate are you proposing for SUPI in LAN mode? Is this a
> proposal that will be aired soon? I thought we had gone well past the
> cutoff date for new proposals in P802.3ae.
>
> Best regards,
> Rich
>
> > Cheers,
> >
> > Paul
> >
> > At 09:38 PM 12/1/2000 -0800, Rich Taborek wrote:
> >
> > >Paul,
> > >
> > >I have to respectfully disagree that the XSBI is in any way a required
> > >interface of the 10GBASE-LW4 PHY. As you clearly indicate in D2.0,
> > >Clause 53, Figure 53-1, the XSBI is an optional interface for the
> > >10GBASE-LW4 PHY. Figure 53-1 clearly illustrates that the WIS may be
> > >directly attached to the LW4 PMA.
> > >
> > >I don't understand your point of suggesting the extraneous insertion of
> > >a high pin count, extraneous parallel interface (the XSBI) into the
> > >10GBASE-LR4 PHY. Is there some specific benefit or purpose of this
> > >suggestion?
> > >
> > >I am very confused by your second paragraph where you mention as SUPI
> > >"extension" to allow WWDM attachment to the LAN 64b/66b encoder over
> > >XSBI.
> > >
> > >1) All of the aforementioned elements, with the exception of the XSBI,
> > >are required for the 10GBASE-LW4 PHY;
> > >
> > >2) SUPI is a PMA which is only applicable to the 10GBASE-LW4 PHY. It is
> > >not applicable to either the 10GBASE-LR4, 10GBASE-XR, or 10GBASE-XW
> > >PHYs.
> > >
> > >Here are my specific questions:
> > >
> > >A) What is a SUPI extension?
> > >
> > >B) How does a SUPI extension enable the use of the XSBI to "all optical
> > >modules"?
> > >
> > >C) How would the XSBI be used as an optical interface for the
> > >10GBASE-LR4 PHY? Do you have a proposal to do this?
> > >
> > >D) What is the advantage of using the 64b/66b PCS for the 10GBASE-LR4
> > >PHY? Do you have a proposal to do this?
> > >
> > >E) What specification/Clause specifies SUPI electrical operation,
> > >specifically the channel model, jitter budgets, transmit eye and
> receive
> > >eye? I am not aware that this is covered by any clause including Clause
> > >47. In essence, the lower line rate advantage of SUPI is offset by
> > >undesirable characteristics which significantly increase jitter such as
> > >very long and unpredictable run lengths. It appears that no such
> > >specifications exist.
> > >
> > >Best Regards,
> > >Rich
> > >
> > >--
> > >
> > >Paul Bottorff wrote:
> > > >
> > > > Ted:
> > > >
> > > > I agree, the two demarcations are XSBI and XAUI. SUPI is not an
> interface
> > > > as standardized by IEEE. Instead SUPI supports the 10GBASE-LW4 by
> attaching
> > > > WWDM to and XSBI interface. The 10GBASE-LW4 WWDM is supported
> attached to
> > > > XSBI using SUPI, while the 10GBASE-LR4 WWDM can only be supported by
> > > > implementing clause 48 and cannot attach to XSBI.
> > > >
> > > > Though late for this discussion, it is possible to extend SUPI to
> allow
> > > > WWDM attachment to the LAN 64b/66b encoder over XSBI. This has three
> > > > advantages: 1)it means all optic modules and modes could be
> supported using
> > > > a XSBI interface; 2)the transmission frequency of the 10GBASE-LR4
> would be
> > > > reduced from 3.125 MHz per lane to 2.578125 MHz per lane; 3)we end
> up with
> > > > a single 64b/66b PCS layer for all interfaces. Using this scheme
> XSBI could
> > > > support all optic module types. In addition, the XAUI option would
> also
> > > > support all types.
> > > >
> > > > Cheers,
> > > >
> > > > Paul
> > > >
> > > > At 01:35 PM 12/1/2000 -0800, Speers, Ted wrote:
> > > >
> > > > > I realize this is an implementation question, but is there
> a
> > > likely
> > > > >demarcation point between the transceiver module and the MAC. It's
> > > hard to
> > > > >sort out a lot of these discussions (LSS, XAUI/SUPI, etc.) without
> > > have this
> > > > >in proper context.
> > > > >
> > > > > Possible points of demarcation would seem to be either the
> > > XAUI or
> > > > >the XSBI interfaces. I've seen presentations suggesting both.
> > > > >
> > > > > It seems that a break at the XSBI would offer the most in
> > > terms of
> > > > >end-user flexibility and economies of scale for both the users and
> the
> > > > >vendors ... with one exception, 10GBASE-X4 would be left out in the
> cold
> > > > >because, as far as I can tell, there is no way to implement the
> standard
> > > > >across an XSBI.
> > > > >
> > > > > Ted Speers
> > > > > Strategic Marketing
> > > > > Actel
> >
> > Paul A. Bottorff, Director Switching Architecture
> > Enterprise Solutions Technology Center
> > Nortel Networks, Inc.
> > 4401 Great America Parkway
> > Santa Clara, CA 95052-8185
> > Tel: 408 495 3365 Fax: 408 495 1299 ESN: 265 3365
> > email: pbottorf@xxxxxxxxxxxxxxxxxx
>
> -------------------------------------------------------
> Richard Taborek Sr. Phone: 408-845-6102
> Chief Technology Officer Cell: 408-832-3957
> nSerial Corporation Fax: 408-845-6114
> 2500-5 Augustine Dr. mailto:rtaborek@xxxxxxxxxxx
> Santa Clara, CA 95054 http://www.nSerial.com