Re: [802.3BA] 5 Criteria mod to support 40 G on SMF
Brad,
Good points.
40GE 10km SMF is similar to 10GE 10km SMF (rather then 100GE 10km SMF)
with respect to having good link budget margin (please see the Technical
Feasibility presentations to the 40GE Ad Hoc from Finisar, OpNext and
Sumitomo/Excelight.) The cost saving associated with a shorter reach
(for example 3km or 4km) is modest. To justify a shorter reach vast
majority of applications (>90%?, >95%?) would need to be shorter then
that reach.
Unlike at 100GE, there is no proposal for a 40GE 40km reach that would
cover the 40GE shorter reach to 10km applications. End users that need
40GE longer then the shorter reach objective (for example above 4km)
would have no solution available to them.
Chris
-----Original Message-----
From: Brad Booth [mailto:bbooth@AMCC.COM]
Sent: Friday, February 15, 2008 11:38 AM
To: STDS-802-3-HSSG@LISTSERV.IEEE.ORG
Subject: Re: [802.3BA] 5 Criteria mod to support 40 G on SMF
Chris,
Personally, I have no disagreement with the objectives the task force
has created. There was excellent data presented to support a 100G, 10
km SMF objective. There was also a large amount of information that was
presented to support the addition of 40G to the project to support the
server market. A huge amount of work and effort went into developing
the task force's current set of objectives.
In relationship to a new objective such as 40G, 10 km SMF, now is the
time to disagree or express an opinion whether it be on the addition of
the objective or the reach target or market potential, etc.
Thanks,
Brad
-----Original Message-----
From: Chris Cole [mailto:chris.cole@FINISAR.COM]
Sent: Friday, February 15, 2008 11:02 AM
To: STDS-802-3-HSSG@LISTSERV.IEEE.ORG
Subject: Re: [802.3BA] 5 Criteria mod to support 40 G on SMF
The 10km reach for Higher Speed data center and campus applications has
been periodically challenged based on individual or corporate
perspectives that see a shorter reach as better. This has continued
despite lengthy discussions and study by the HSSG of reaches used for
past data rates, reach distributions, and end user needs.
Last year, I was one of those that held the view that a shorter reach
may be a better objective. This was based on finding a cost difference
between 100G 10km SMF PMD versus 100G 3km or 4km PMD. I also had an
impression that a shorter reach would be sufficient for most (>90%) data
center links. That impression was formed based on interaction with a
small number of end users.
At the November HSSG meeting, three optics vendors; Finisar, OpNext and
Sumitomo/Excelight brought the PMD cost difference to the attention of
the HSSG and requested another re-examination of the 10km reach
objective. Despite the 10km reach having been formally adopted as an
objective, and despite approval to move forward to Task Force, the chair
and the HSSG took the request seriously and devoted considerable time
and effort to again study this issue.
This was done at the November meeting, in reflector correspondence,
Reach Ad Hoc presentations and discussions, and efforts by multiple
individuals to get end user information through outside forums. We also
presented a very simple methodology to approximately quantify reach
objective cross-over based on relative PMD costs and reach
distributions.
The outcome of this study was that 10km reach objective was again found
to meet the needs of majority of end users. To date, none of the
presented PMD cost differences between 10km and 3/4km justify making a
change to the 10km reach objective given this reach distribution. It is
true that we would feel more comfortable if we had a definitive,
comprehensive, industry wide study of data center/campus reach
distributions. Unfortunately, such a study does not appear to exist and
our decision has to be based on many partial information sources.
10km is the best reach objective based on what we know today. There is
no need to feel uneasy that it was selected based on less then perfect
information; that is the nature of such decisions.
Those that hold the perspective that a shorter reach is better need to
bring in solid industry wide data and/or significant end user support.
They further need to show an analysis of how their reach distribution
data combined with PMD cost differences justifies changing the reach
objective. Continuing to simply express a different opinion or
dissatisfaction is not helpful at this point.
Chris
-----Original Message-----
From: Brad Booth [mailto:bbooth@AMCC.COM]
Sent: Thursday, February 14, 2008 1:00 PM
To: STDS-802-3-HSSG@LISTSERV.IEEE.ORG
Subject: Re: [802.3BA] 5 Criteria mod to support 40 G on SMF
Steve,
I agree that this does change things up. 40G was positioned as a server
interconnect, and 100G as the network interconnect. The overlap with an
existing 100G objective blurs the distinct identity making it hard to
differentiate the broad market potential for either objective.
If the 40G SMF option was being targeted at a 2 km reach, that would
make sense considering that's a typical campus area network. If someone
then wanted to use that to do 10 km, that would be their option and
considered outside the scope of the standard.
Thanks,
Brad
-----Original Message-----
From: Swanson, Steven E [mailto:SwansonSE@CORNING.COM]
Sent: Thursday, February 14, 2008 1:08 PM
To: STDS-802-3-HSSG@LISTSERV.IEEE.ORG
Subject: Re: [802.3BA] 5 Criteria mod to support 40 G on SMF
Howard etal,
I could agree that this proposal may address what needs to change in our
response to the Economic Feasibility criterion to add the 10km SMF
objective at 40G. However, I think this change goes against the basis of
the decision that we took in July 2007 for considering 40G in the first
place, i.e., segmenting the server and computing applications from the
network aggregation applications. Now the network aggregation space
includes two solutions - 40G and 100G. This seems to place a burden on
us to also re-evaluate Distinct Identity criterion and calls into
question the Broad Market Potential criterion. In reviewing some of the
presentations leading up to our decision to include 40G, I note some of
the concerns expressed then that I think are now back on the table:
* "Fragmentation of R&D efforts (lack of critical mass on either
40G or 100G initially) - two rates will ultimately force component and
equipment vendors to support BOTH."
* "Requires the industry to develop 2 x MACs, 2 x PCS chips, 2 x
PMA (serdes) chips, 2 x N PMDs"
* "Industry confusion on "application versus rate" - Distinct
Identity does not just mean "Is there anything else exactly like this?"
but also "Is there sufficient difference between this and available
alternatives to justify the effort?"
* "Interoperability concerns (some vendors elect to implement 40G
initially, whereas others implement 100G)"
* "Do we now have a triple rate required (40G LAN, 40G WAN, 100G)
since 40G SMF solutions already exist?"
I do not have any proposed changes at this point regarding the BMP or DI
criterion but wanted to express an opinion that these criterion should
be re-evaluated. If others do not share that view, then on we go.
Steve Swanson
Corning Incorporated
-----Original Message-----
From: Howard Frazier [mailto:hfrazier@BROADCOM.COM]
Sent: Thursday, February 07, 2008 6:21 PM
To: STDS-802-3-HSSG@LISTSERV.IEEE.ORG
Subject: [802.3BA] 5 Criteria mod to support 40 G on SMF
Dear members of the IEEE 802.3ba Task Force,
I have reviewed our approved set of 5 Criteria responses
http://www.ieee802.org/3/ba/PAR/HSSG_5C_0707.pdf
looking for any material that will need to be changed in the event we
adopt an objective to support 40 Gb/s operation on 10 km of single mode
fiber. In my opinion, the responses will remain valid and complete, with
one exception.
On page 6 of the above referenced file, in our response to the Economic
Feasibility criterion, we state:
Presentations indicate that for the server market and
computing applications the optimized rate to provide
the best balance of performance and cost is 40 Gb/s.
For the network aggregation market and core networking
applications, the optimized rate offering the best
balance of performance and cost is 100 Gb/s.
If we adopt a 40 Gb/s SMF objective, then this response should be
modified along the lines of:
Presentations indicate that for the server market,
computing applications and some cost-sensitive
aggregation applications, the optimized rate to provide
the best balance of performance and cost is 40 Gb/s.
For the network aggregation market and core networking
applications, the optimized rate offering the best
balance of performance and cost is 100 Gb/s.
The change being the insertion of the words "some cost-sensitive
aggregation applications" in the first sentence.
I think that our previously approved responses for Broad Market
Potential, Compatibility, Distinct Identity, and Technical Feasibility
will not require any change in the event that we adopt an objective for
40 Gb/s operation on SMF. I think that the proponents of the new
objective will be able to readily demonstrate this.
If you think I have over looked something else that might need to be
changed, please speak up, and please provide a proposed change.
Howard Frazier
Broadcom Corporation