To add to Paul’s comments I’ll
put some actual power dissipation numbers next to this for perspective. I
believe existing 10GBASE-T solutions are in the neighborhood of 10 W, with newer
implementations going down to 5-6 W. Someone from the 10GBASE-T community
can correct me if I’ve got it wrong, as I’m not an expert in this
area. So moving to ~25W in the near future for 4 10GBASE-T ports. This
is compared with a 40G QSFP at 3.5W or 4 SFP+ modules at 1W apiece.
So at least for the next couple of years, the
big factor in power dissipation for a 4x10G LAG vs. 40G solution would likely
be whether it’s done in copper or fiber, not whether it’s a 4x10 LAG
or 40G solution. Where the power dissipation numbers go from there, like
Paul said, depends on a lot of factors and is up for (no doubt spirited) debate.
Scott
>>Scott Schube
>>Strategic Marketing Manager
>>Intel Optical Platform Division
>>(510) 578-5815
>>scott.a.schube@intel.com
From: Paul Kolesar
[mailto:PKOLESAR@SYSTIMAX.COM]
Sent: Tuesday, April 10, 2007 3:45
PM
To:
STDS-802-3-HSSG@listserv.ieee.org
Subject: Re: [HSSG] 40G MAC Rate
Discussion
First, let me apologize for misspelling Scott Kipp's
last name in my previous note. Sorry Scott.
Mikael,
I
think you have touched on the key issue facing 10GBASE-T: power dissipation. While
we can look forward to lower power consumption with future IC feature
reductions, these will be muted by the need for multiple 10GBASE-T PMDs in a
LAG or other parallel scenario. Such power reductions will also enable
lower power consumption by the optical PMDs, although by relatively smaller
power reduction ratios. In my estimation it will be difficult, or far
into the future, to deliver a 4x 10GBASE-T device that dissipates the
equivalent dissipation of 40G parallel optical transceivers.
The
cost of the copper cabling will also be 4x higher than for a single 10GBASE-T
link. So here it would loose its most of its competitive edge.
The
relative PMD cost factors are unknown, since I am not aware of 10GBASE-T ports
in the market at this time. But it is very likely that once 10GBASE-T
become highly integrated, they will compete with the parallel optical
transceivers where their application spaces overlap. It is unclear if
that overlap will include 40G applications, since LAG deployments will fade as
the single PMD solutions become available.
Time
appears to be one of the variables with the greatest influence on the optimal
choice. Will that play best to copper or fiber? Somebody with a
better crystal ball than mine on the issues outlined above needs to weigh in.
Regards,
Paul Kolesar
CommScope Inc.
Enterprise® Solutions
1300 East Lookout Drive
Richardson, TX 75082
Phone: 972.792.3155
Fax: 972.792.3111
eMail: pkolesar@commscope.com
|
Mikael Abrahamsson
<swmike@SWM.PP.SE>
04/10/2007 03:04 PM
|
Please
respond to
Mikael Abrahamsson <swmike@SWM.PP.SE>
|
|
|
To
|
STDS-802-3-HSSG@listserv.ieee.org
|
|
cc
|
|
|
Subject
|
Re: [HSSG] 40G MAC Rate Discussion
|
|
On
Tue, 10 Apr 2007, Ali Ghiasi wrote:
> I know there is readily available solution
based on QSFP that could do
> 40G copper with twin-ax cables. Copper
solution are very attractive for
> sub-10m as they offer order of magnitude cost
advantage. Longer reach
> cable not as attractive as you have to use
heavier gage which increase
> the cable size, weight, cost, and the
complexity of receiver increases.
Well, my idea was to compare 4*10GE LAG and
whatever 40GE solution is
being proposed. I understand now I should have
written this outright,
sorry.
My thoughts were that operational cost and CAPEX
of doing 4*10GE LAG using
10GBASE-T for servers would be quite a lot lower
than any new proposed
40GE solution, since I only saw coaxial and fiber
proposed for 40GE
physical media.
Or have I misunderstood 10GBASE-T when it comes to
power and reach (100M
over CAT7 or ~50 on CAT6)? Does it require a lot
more power than fiber
based solution?
From a pure end-user perspective CAT6-7 based
cabling has the lowest
operational cost (physical only) since it's
abundant and there is a lot of
experience with that kind of cabling. But perhaps
it uses a lot more power
than the other proposed solutions for 40GE?
--
Mikael Abrahamsson email:
swmike@swm.pp.se