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

Re: What is 802.3ae WAN-PHY?




Osamu,

Please understand that I am under certain levels of Non-Disclosure for some
of your issues.  The real consideration for P802.3ae TF is that at 10Gb, the
P802.3ae WAN compatible PHY will have complete occupancy of any fiber or
wavelength that it rides on.   Any muxing that is done, is done at the
wavelength level within DWDM, not at the data signal level.  This is
recognized as changing the rules for needing common transmission clocking,
also known as "Stratum" clocking.  While the transmission systems themselves
are still required to provide for close clock transfer tolerance, the full
wavelength data originating systems are no longer under that restriction.
There is, or soon will be, one or more contributions to the other standards
bodies regarding this.

Subrates of SONET/SDH are still multiplexed at the TDM data signal level.
The data sources for these still require the same level of common
transmission clocking as always.  For that reason, EOS at OC3/STM1 through
OC189c/STM63c still require loop timing off of the Stratum clocked
transmission systems.  Any transmission systems that provide 802.3 Ethernet
interfaces, at any rate, must buffering/bridging functions that provide for
the adjustment between +/- 100ppm clocking of Ethernet and the +/- ~4ppm
clocking of the Stratum timed transmission systems.  Ethernet data switches
and routers that implement subrate interfaces, such as the ITU PRC EOS
proposal will require the normal loop timing and close clock tolerance of
the existing SONET and SDH systems.  You need not have any concerns.  For
normal multiplexed SONET and SDH systems, I do not foresee any proposals to
ANSI or ITU to change these timing requirements.

As you can see, there is a major distinction between the proposed
requirements of P802.3ae and the normal TDM multiplexed SONET and SDH
transmission systems.  Any Ethernet data switch and router interfaces that
are at SONET/SDH  multiplexed subrate speeds will continue to be expensive
compared to the proposed P802.3ae WAN compatible PHY interface.   I am sure
that there are others in the TF with transmission systems experience that
can confirm these distinctions.

Thank you,
Roy Bynum



----- Original Message -----
From: Osamu ISHIDA <ishida@xxxxxxxxxxxxxxxxxxx>
To: Roy Bynum <rabynum@xxxxxxxxxxxxxx>; David Martin
<dwmartin@xxxxxxxxxxxxxxxxxx>
Cc: <stds-802-3-hssg@xxxxxxxx>
Sent: Sunday, April 09, 2000 2:06 AM
Subject: Re: What is 802.3ae WAN-PHY?


>
> Hi Roy, Hi Dave,
>
> Thank you for your feedback.  Your comments are helpful for me to
> understand what was going on in HSSG and T1, since I have joined
> HSSG since the Dallas meeting (Jan. 2000) and I have no liaison
> to T1.
>
> Before discussing in detail, I think we had better be aware once
> again that here is a LAN community, and keep in our mind that we
> should be much careful to use WAN terminology.  I use the
> following definitions in this mail; please let me know if there
> is any inconvenience to you.
>  --------------- --------------- ------------ ----------- ------
>  expression      Overhead access CLK Accuracy CLK Jitter  Std.
>  --------------- --------------- ------------ ----------- ------
>  SONET           (almost) full   < +/-20ppm   SONET spec. T1.105
>  SONET-compliant reduced         < +/-20ppm   SONET spec. T1.416
>  SONET-lite      much reduced    < +/-100ppm  ??????????  ??????
>  --------------- --------------- ------------ ----------- ------
>  Here I assume that SONET-lite equals WAN-PHY with SONET framer
>  proposed in 802.3ae.
>
> I think we can agree on the following two facts;
>
>  (1) There is no SONET-lite standard at present.  All SONET or
>      SONET-compliant equipment at present, including the emerging
>      non-muxing LTE, should meet +/-20ppm CLK accuracy and SONET
>      jitter spec.
>
>  (2) Pointer manipulation mechanism in SONET framer itself has
>      the potential to accommodate the CLK difference up to +/-320ppm.
>
> According to Roy, there are some discussions on relaxing the SONET
> CLK accuracy for some specific applications in T1 or ITU-T.  That's
> news to me; please let me know where I can get the more information
> about the discussion.
>
> I have no authority to decide how NTT reacts in ITU-T, but I will
> strongly recommend my colleagues who often attend the ITU-T expert
> meeting not to approve such standard proposals.  The reasons are;
>
> (1) I know we have at least one SDH system that raises an alarm for
>     too many pointer re-writing to indicate wrong CLK accuracy.
>
> (2) It can not be connected to SONET regenerators.  This implies
>     that it is not a SYNCHROUNOUS system any more.  I see no reason
>     to define 'plezioisosynchronous' system with SONET framer which
>     is optimized for SYNCHRONOUS systems.  Note that we have a
>     serious trade-off between CLK accuracy and jitter transfer in a
>     SONET regenerator (you can read it in SYNCHRONUS transport systems).
>     I think this is what Erik has pointed out in his recent comment.
> http://grouper.ieee.org/groups/802/3/10G_study/email/msg02157.html
>
> (3) I'm afraid that it would nominate the client path; only OC-192c
>     or some small variations could be carried through non-muxing LTE.
>     We will lose flexibility to accommodate any combination of OC-1,3,
>     12,48,192.  Or we require very heavy Line terminating circuit to
>     re-write up to 192 pointers independently.
>
> Above consideration tells me that, if we use the SONET framing,
> we had better keep +/- 20ppm accuracy to be compliant to SONET.
> Then it can be connected directly to existing SONET regenerators and
> WDM active transponders.  We don't need to pay the SONET-framing
> charge twice for WAN-PHY and ELTE.  And again, I don't believe
> SONET regenerator for 10GbE is a good choice unless you have already
> invested in it.
> http://grouper.ieee.org/groups/802/3/10G_study/email/msg01869.html
>
> Best Regards,
> Osamu
>
> At 14:39 00/04/07 -0500, Roy Bynum wrote:
> http://grouper.ieee.org/groups/802/3/10G_study/email/msg02169.html
> >
> > Osamu,
> >
> > "SONET Lite" is a term that has been used for some time to mean a
reduced
> > overhead requirement customer drop interface.  T1.416-99 defines the
> > required and optional byte usage for "SONET Lite".  The
plezioisosynchronous
> > timing standard is currently being worked on.  It should be a full
standard
> > definition very soon.  In this regard, the "ELTE" is acutely a special
case
> > non-muxing LTE that is part of the T1X1 standards.  The SONET test sets
that
> > are used for POS today already have to be adjusted to allow for wider
clock
> > and jitter tolerance.
> >
> > Thank you,
> > Roy Bynum
> >
> > ----- Original Message -----
> http://grouper.ieee.org/groups/802/3/10G_study/email/msg02164.html
>
> At 15:42 00/04/07 -0400, David W. Martin wrote:
> http://grouper.ieee.org/groups/802/3/10G_study/email/msg02170.html
> >
> > Osamu,
> >
> > Our WAN PHY proposal uses the traditional +/-100ppm clock tolerance spec
> > of Ethernet.  An ELTE would have a matching WAN PHY interface. The SONET
> > side of the ELTE will have an ITU/ANSI interface. Within the ELTE the
clock
> > domain adaptation is performed using traditional SONET pointer
processing.
> > Previous material has shown this is feasible. Please refer to
> >
> >
http://grouper.ieee.org/groups/802/3/10G_study/public/july99/bottorff_1_0799
.pdf
> >
> > for more background. Thank-you.
> >
> > ...Dave