Matt,
In reviewing your 40GE SMF cost email I am
puzzled by the conclusions you reached with respect to ICs and integration of
10G PHYs. At the May HSSG meeting, Howard Frazier (Broadcom) gave a
presentation titled “Evolution from 10G to 40G & 100G”.
http://www.ieee802.org/3/hssg/public/may07/frazier_01_0507.pdf
The main theme of the presentation is
evolution of single 10G PHYs => Quad 10G PHYs (i.e. 4x10G ICs). On almost
every page of the presentation, Howard shows Quad 10G PHYs supporting copper
and optical interconnect. That presentation has supporters from the following
companies: Intel, Sun, NTT, CommScope, Nortel, Avago, Juniper Networks, OFS,
and IBM. Very low cost Quad 10G PHY is critical to the economic feasibility of
40GE server interconnect. The clear implication from the presentation is that
server volume will drive down the cost of 4x10Gs ICs to much lower level then optical
interconnect volume can drive down 40G serial IC cost.
Will it be possible to get more details of
your company’s IC cost analysis to see why it leads to different
conclusions then for the server industry?
It would also be helpful to get confirmation that your proposal does not
support 100m OM3 MMF interconnect. I assume that you are not suggesting that
40G serial can be used over MMF. In the below January presentation to the
802.3ba Task Force, Adam Carter and Alessandro
Barbieri proposed 40GE over legacy duplex OM3 MMF as a
requirement to support access to distribution in legacy data centers with cable
infrastructure designed for 10 GbE.
http://www.ieee802.org/3/ba/public/jan08/barbieri_01_0108.pdf
Did you hear that this is not an important
requirement from various end users (I assume that’s who “various
folks” are in your email)?
Thank you
Chris
From: Matt Traverso
[mailto:matt.traverso@gmail.com]
Sent: Saturday, February 16, 2008
10:55 AM
To: Chris
Cole
Cc: STDS-802-3-HSSG@listserv.ieee.org
Subject: Re: [802.3BA] 5 Criteria
mod to support 40 G on SMF
Chris Cole: "If the 802.3ba
Task Force decides to add a 40GE SMF reach objective, the
additional workload to write the standard will be moderate."
*** Agreed.
Chris Cole: "Just as
development of a 40GE SMF PMD will highly leverage the existing 10GE
technology, so will specification writing highly leverage 10GE-LR and LX4
standards. There appears to be a consensus on the approach, so no technical
controversies are expected."
*** Unfortunately I cannot agree. My company and others are intimately
aware of the challenges with the parallel approach. Based upon the volume
ramp comments/perspectives I have been hearing from various folks on a 40GbE
SMF solution it seems that clear path to minimize the bit/s/km costs should be
optimized for 2012. Based on this data point, we hold a strong opinion
that the 40GbE SMF PMD should be based on serial 40G optics rather than the
4lambda approach.
When LX4 was first dreamed up many folks claimed that it would be lower cost
than 4x2.5G SFP's because all the IC technology existed and the optics would be
integrated. As it turned out that has never materialized. The packaging and yields
costs were much higher than anyone anticipated. Also, the integration of IC's
didn't really happen because IC vendors could not justify making the investment
for such a small volume in the first couple of years.
In undertaking our analysis for 40G serial vs. 4x10G, we have tried to give the
4x10G the benefit of the doubt, but by 2012 it seems that the costs are clearly
shifting in favor of the serial approach.
thank you
--matt traverso
Opnext
On Feb 16, 2008 9:39 AM, Chris Cole
<chris.cole@finisar.com>
wrote:
John Jaeger is exactly right in underscoring that most important in the
40GE SMF reach objective discussion is coming to a quick decision. The
suggestion to make this decision at the March meeting is a good one.
If the 802.3ba Task Force decides to add a 40GE SMF reach objective, the
additional workload to write the standard will be moderate. Just as
development of a 40GE SMF PMD will highly leverage the existing 10GE
technology, so will specification writing highly leverage 10GE-LR and
LX4 standards. There appears to be a consensus on the approach, so no
technical controversies are expected.
If the 802.3ba Task Force decides not to add a 40GE SMF reach objective,
a likely scenario will be as described by Brad Booth
in his last
reflector email. Specifically, optics vendors will develop a
non-standards based PMD to satisfy market demand. The development of a
40GE SMF PMD leverages the huge investment made in 10GE technology, so
makes it desirable from an ROI perspective. This is not the preferred
path, but one that will likely be taken if there is no standard because
of favorable economics.
Chris
P.S. For those that have not gone back to nicholl_01_0507, I suggest you
take a few minutes to take a look. It is almost spooky how accurately
the presentation outlines what is happening.
-----Original Message-----
From: John Jaeger [mailto:jjaeger@IEEE.ORG]
Sent: Friday, February 15, 2008 12:39 PM
To: STDS-802-3-HSSG@LISTSERV.IEEE.ORG
Subject: Re: [802.3BA] 5 Criteria mod to support 40 G on SMF
> -----Original Message-----
> From: Alessandro Barbieri
(abarbier) [mailto:abarbier@CISCO.COM]
> Sent: Friday, February 15, 2008 11:31 AM
> To: STDS-802-3-HSSG@LISTSERV.IEEE.ORG
> Subject: Re: [802.3BA] 5 Criteria mod to support 40 G on SMF
>
<<snip>>
>
> 40G is absolutely incremental as it is a path to upgrade 10G links for
> the more cost driven part of the market. Hadn't we had 40G, most users
> would have continued happily with nx10G links.
>
Now there's an idea I could get behind (again). Happy 4x10G LAG
customers
sounds good to me too. But that would be back tracking the group even
further.
Hope that no one is really surprised that this discussion is taking
place -
as this exact scenario was discussed on multiple occasions last March,
April
and May. It was only a matter of when it would take place, not if.
Gary's
presentation in Geneva
and the 28 supporters who signed onto it clearly
spelled this out
http://www.ieee802.org/3/hssg/public/may07/nicholl_01_0507.pdf
.
So since we are here, my hope is that the Ad-hoc presentations & calls
continue to get the data out for discussion, the reflector dialog
continues
to debate the more interesting points, and that we get a complete airing
and
discussion of the topic in the March meeting so that with a decision
either
way sooner rather than later, we can all focus our efforts on the task
at
hand; proposals and baseline selections for a draft standard.
-john