Dan,
I believe Chris makes some great points, and your answer highlights some
of my heartburn about the process*.
<Chris wrote>:"10G DFB lasers vs. 40G EML. How do you determine their
relative costs in a perfect world?"
<Dan wrote>: Without being glib, I would trust you and your peers to
come up with the relative cost.
All of the subject matter experts enter into the IEEE with similar yet
different tools at their disposal such that it is often difficult to
come to agreement on even component relative costs. Let me point out to
somewhat recent examples:
First, in 802.3ba, there was a debate amongst various VCSEL
manufacturers around the spectral width parameter. According to some it
was "free"; according to others this parameter nearly defined the cost
of the VCSEL. Of course each party was correct for their particular
VCSEL structure. You might recall that this debate received some
attention as there was a push to extend the 100GBASE-SR10 reach and
narrowing the spectral width would help to extend the supportable reach
especially for OM4.
Second, also in 802.3ba, Chris Cole and I engaged in a multi-meeting
cycle debate on the merits of LAN-WDM versus CWDM wavelength spacing for
100GBASE-LR4. I know that in the case of my former employer and
affiliation that CWDM would have represented both a near term and long
term component power/cost improvement over LAN-WDM. However, I am
fairly sure that for Chris and his employer/affiliation that LAN-WDM
represented the optimum near& long term solution given the technologies
they possessed. Again, we were both right with the tools we had in
place, and our committee made a choice. Even as the "loser" in that
particular issue, I believe that it is better that we made a choice
which is optimal for at least some to drive interoperability and network
connectivity forward.
My suggestion going forward: Architecturally describe how a particular
proposal represents a cost/power/density advantage for the particular
application. I'm working toward a contribution to this effect.
*Note: I do believe the IEEE process that is in place with the 5
criteria and the rules we have in place are very very good; I'm just
pointing out that they are not perfect. If I am able to think up some
better rules/processes I promise to float them by David Law and the rest
of the experienced folks in the IEEE SA management team.
aloha
--matt
-----Original Message-----
From: Daniel Dove [mailto:ddove@xxxxxxx]
Sent: Friday, February 03, 2012 4:07 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Minutes Uploaded - Forward Direction
Hi Chris,
I take your point regarding what to compare. ASICs, lasers, fibers, test
and assembly can also be compared by an expert in the field. Other
experts in the field can assess the comparison and nod their heads.
Others in the room can trust their judgment and we have consensus.
"10G DFB lasers vs. 40G EML. How do you determine their relative costs
in a perfect world?"
Without being glib, I would trust you and your peers to come up with the
relative cost.
Lets see what David has to say.
Dan
On 2/3/12 3:55 PM, Chris Cole wrote:
Dan,
I agree with you that this requires a thoughtful offline discussion
and
well articulated guidelines.
The perfect world you describe is unfortunately far removed from
reality. PCB type, number of components, etc are not what drives cost
differences. It is ASICs, lasers, fibers, test, and assembly. One
example comparison from 802.3ba is illustrative: 10G DFB lasers vs.
40G
EML. How do you determine their relative costs in a perfect world?
Chris
-----Original Message-----
From: Daniel Dove [mailto:ddove@xxxxxxx]
Sent: Friday, February 03, 2012 3:48 PM
To: Chris Cole
Cc: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Minutes Uploaded - Forward Direction
Hi Chris,
I also do not want to rely upon my own interpretation and have
requested
an opinion from David Law.
In my perfect world, someone would present an analysis on relative
cost
of X vs Y using size, PCB type, thickness, number of comparable
components, processing differences, die size, power, and the like to
come to a relative cost between X and Y.
They would present that at the meeting. People would either nod their
heads and agree, or they would challenge assertions and perhaps not
come
to consensus. Consensus does not mean everyone having the same
numbers,
it means everyone nodding their head and agreeing that the analysis is
"close enough for engineering". Consensus might be that the presenter
was off on their numbers, and more presentations would have to come in
to get consensus.
Once that had been done, others could compare Z's relative cost to
either X or Y and the group would now have a feel for how X, Y and Z
stack up.
I am just not comfortable taking "cost" based on market information.
That is not cost IMHO.
Lets put a damper on this discussion until David Law can provide his
opinion.
Dan
On 2/3/12 3:33 PM, Chris Cole wrote:
Steve,
You have fully captured the difficult problem we face in trying to do
due diligence on competing proposals.
If we use completely sanitized solution A and solution B costs which
cannot be referenced to more generally available data, then we are
back
to "liar's poker" or more generously to "he said, she said". In that
case, there is no way to challenge the cost assertions of proponents
of
solution A or solution B.
If on the other hand we use more generally available data, so that
there
is objectivity in the cost comparisons, not just unsubstantiated
claims,
then this can always be traced to an actual dollar amount even if
through a tortuous path.
So the concern I raised about not being able to demonstrate 5
Criteria,
as further commented on by Brad, is genuine.
The solution we employed in 802.3ba is to use cost numbers averaged
over
many suppliers and many variants. In practice there is a very wide
variability in the market around that baseline, and the specific
average
number may not even exist. The information is equally available to
everyone so it does not confer an advantage on any party in the
market.
Despite all this, it is obviously not as crisp as not using any
generally available relative cost data.
We are going to have to agree to some more useful guidelines than
have
been given so far if we don't want to rely on ad hoc interpretations
by
the chairs of every email or presentation.
Chris
-----Original Message-----
From: Trowbridge, Stephen J (Steve)
[mailto:steve.trowbridge@xxxxxxxxxxxxxxxxxx]
Sent: Friday, February 03, 2012 2:06 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Minutes Uploaded - Forward Direction
Hi Chris,
I think that a lot of folks on the list are having trouble
distinguishing the difference in the meaning of "price" and "cost",
but
also about the meaning of the word "relative".
If people start describing costs relative to something they can buy
today, and everybody knows the price of what they can buy today,
people
assume that if the same margins were to apply, they can infer a
relative
price. This gets dangerous in that you are essentially deciding "how
big
is a banana", and then using bananas as currency.
The idea of "relative" to my understanding is that you need to
restrict
yourself to "A" vs. "B" comparisons, and not put the universe on a
common scale which amounts to defining a currency which basically
translates to price. So if you have two ways to solve a problem, you
can
compare the relative costs of those two solutions. If you are trying
to
test if a "replacement" PMD is justified, you would want to analyze
the
long term cost of what is inside of the module to see if that looks
to
cost enough less, long term, to be worth building.
What I see people trying to do is to establish this currency where
folks
are asked to compare their own proposal to some baseline (e.g., SR10)
assuming that if everyone uses the same baseline, you can do
arithmetic
and effectively compare the relative costs of solution A vs solution
B,
not by actually comparing them directly against each other, but by
using
some other established currency. I am not sure this is a really
productive exercise either, since I think it would just create a game
of
"liar's poker" with every contributor trying to construct a story
that
puts their solution at the lowest "cost" based on this common
currency.
This assumption is wrong unless you can guarantee that everyone who
does
this is equally thorough, equally honest, equally optimistic (or
pessimistic). The common scale won't help you: you really need to
look
at solutions directly against each other.
Regards,
Steve
-----Original Message-----
From: Chris Cole [mailto:chris.cole@xxxxxxxxxxx]
Sent: Friday, February 03, 2012 2:35 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Minutes Uploaded - Forward Direction
John
Can you explain the difference between the discussion of relative
module
costs in NG 100G OE SG, and the discussion of relative module costs
in
HSSG and 802.3ba including of specific form factors?
Chris
-----Original Message-----
From: John D'Ambrosia [mailto:jdambrosia@xxxxxxxxxxxxxxx]
Sent: Friday, February 03, 2012 1:14 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Minutes Uploaded - Forward Direction
Dan,
Let me clear - do not include me on any discussions regarding price.
That is an inappropriate discussion and I do not want to be involved!
John
-----Original Message-----
From: Daniel Dove [mailto:ddove@xxxxxxx]
Sent: Friday, February 03, 2012 4:11 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Minutes Uploaded - Forward Direction
Seriously, folks;
I am going to officially ask that if you wish to talk about products
and
their relative prices, please do it on a different discussion forum.
Maybe an email thread among friends?
We are here to talk about IEEE standards, PMDs, and their relative
costs.
Dan Dove
Chair, Next Generation 100G Ethernet Study Group