Scott,
I agree that the longer the channel,
the more likely it is that it would be composed of concatenated links.
For example, with my corporate sales data for data center backbone
links (A-D and D-C subsystems) aligns with the survey data from Alan Flatman
to indicate that about 45% of all channels in his survey might be composed
of two concatenated links. See kolesar_01_0308. Such channels
require four MPO connections, which is why there have been repeated requests
to acknowledge this type of topology with in the standard.
While I agree with the use of statistical
models for connection loss, I know that the statistics that Jonathan King
used are probably optimistic with regard to the installed base of MPO connections.
But the loss of the MPO connectors is improving over time, and premium
low-loss MPO offerings now exist from many cabling vendors.
In light of these facts, the key question
becomes how much of the installed base of MPO-based cabling (which is currently
serving multiple 2-fiber applications) will be converted (by removing the
modular fanouts) to cabling that supports array-based applications like
40GbE and 100GbE. If the answer is none, then we don't have to support
the installed base. If the answer is some, or much, or most, then
we need to allocate the proper amount of loss budget to permit them to
operate. Here, a sufficient allocation for 4 installed base MPOs
is about 2 dB. For the newer premium low-loss MPOs a sufficient allocation
is about half that.
I agree that advice on the trade off
of link length with connection loss would be valuable information in the
standard. This is the type of material that belongs in an informative
annex.
Regards,
Paul Kolesar
CommScope Inc.
Enterprise Solutions
1300 East Lookout Drive
Richardson, TX 75082
Phone: 972.792.3155
Fax: 972.792.3111
eMail: pkolesar@xxxxxxxxxxxxx
Scott Kipp <skipp@xxxxxxxxxxx>
08/27/2008 02:21 PM
Please respond to
Scott Kipp <skipp@xxxxxxxxxxx>
To
STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
cc
Subject
Re: [802.3BA] 802.3ba XR ad hoc next
step concern
All,
One topic that should be covered in this discussion of longer XR links
is increased connector loss. If the link is longer, there is a higher
chance that the link goes through additional patch panels and the number
of connectors will likely exceed 4. In addition to having more
connections, the MPO connector typically has considerably more loss as
an LC connector which was the baseline for previous Ethernet optical
links.
If anyone can collect some information on how these 150+ meter links are
layed out it would be very helpful. From my reading of flatman_01_0108,
it shows that 10%+ of access-to-distribution and distribution-to-core
links fall between 100 meters and 300 meters. It would be helpful
to
understand these XR links that must be out there, but many people don't
understand.
I covered connector loss in kipp_01_0108 and recommended having an annex
that specified what the supported link distance would be with 3.0 dB of
connector loss. Maybe 3.0 dB is too high, but 1.5 dB of connector
loss
could easily be exceeded on a 200 meter link that winds its way through
multiple patch panels through the data center.
One criticism of my presentation was that we should have a statistical
model of the connector loss and Jonathon King provided a nice
statistical analysis in king_01_0108 with up to 4 connectors. If these
XR links went through one more patch panel which usually adds 2 more
connections, the likely mean connector loss would be around 1.25 dB and
probably 30-40% of links would exceed 1.5 dB of connection loss. If
the
links went through an additional patch panel and was up to 8 MPO
connections, then the majority of links would have more than 1.5 dB of
connector loss. When you consider that a 150 meter link has higher
fiber attenuation than the 100 meter link, the increased connector loss
could push these links into troubled waters.
It would be good to model the link with 2.0 dB of connector loss as well
as the higher fiber loss to get an understanding of how the link will
work under this likely XR condition.
Thanks,
Scott
-----Original Message-----
From: Swanson, Steven E [mailto:SwansonSE@xxxxxxxxxxx]
Sent: Wednesday, August 27, 2008 1:24 PM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3BA] 802.3ba XR ad hoc next step concern
Mikael,
According to data presented over the last several meetings, I would
offer the following:
100m over MMF: 1x
150-200m over MMF: 1.2x
10km over SMF 8-12x
The reason people still invest money in multimode fiber is because it is
the most cost effective optical choice for short lengths (< 300m)
Best regards,
Steve
-----Original Message-----
From: Mikael Abrahamsson [mailto:swmike@xxxxxxxxx]
Sent: Wednesday, August 27, 2008 1:34 PM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3BA] 802.3ba XR ad hoc next step concern
On Wed, 27 Aug 2008, Dove, Daniel wrote:
> Perhaps you can provide a chart that shows the distribution of lengths
> and the relative costs associated with them, and we can use that as
> the basis for further understanding? I know that this data has been
> presented in various presentations, but a nice succinct chart would
be
> helpful.
I'm also interested in the cost difference between 100m, cost for an MM
optic that'll do 150-200m, and if the optic that does 10km over SMF is
applicable for the 200m MM case (like it is with 10GBASE-LX4 and
1000BASE-LX for instance), and if it is, what's the relative cost with
that included.
On a personal note, I still don't understand why people invest a lot of
money in MM cabling when SM is around :P It's obvious that to use MM
with price-efficient optics, each hike in speed seems to shorten the
reach in MM fiber.