Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [802.3_100GNGOPTX] Emerging new reach space



Alan,

 

Thanks for raising this conflict of inputs.  I think it reflects the lack of knowledge that the industry has about what is going on.  I read lots of analyst reports and when I try and correlate the results, the numbers don’t correlate.

 

Let’s look at the analysis in more detail:

Gartner - 2011

Rack Range

Square-Foot Range

2010

2010

2015

2015

Rack/Computer Room

2-25

Up to 750

488494

95.60%

436441

94.80%

Midsize DC

26-100

Up to 3,000

15669

3.07%

16013

3.48%

Enterprise DC

101-500

Up to 15,000

5668

1.11%

6240

1.36%

Large DC

500+

>15,000

1110

0.22%

1361

0.30%

Total

 

 

510941

 

460055

 

Gartner didn’t give a range for the number of servers

IDC - 2006

Server Range

Square-Footage of Raised Floor

2003

2003

2008

2008

Small Data Center

351-503

15,000

2200

47.41%

2300

46.37%

Medium Data Center

1,504 - 1,705

20,000

1450

31.25%

1600

32.26%

Large Data Center

2,006 - 2,507

35,000

900

19.40%

950

19.15%

Very Large Data Center

>25,008

>100,000

90

1.94%

110

2.22%

Total

 

 

4640

 

4960

 

IDC doesn’t give a contiguous range of servers or square footage.  Where would you categorize a data center with 1,000 servers?  Their data is dated by 5 years now too.  The odd numbers for the range of servers includes volume servers and high end servers.

 

Besides having 4 categories, not much else is the same in these two surveys.  Here’s a summary of some differences:

 

Number of data centers and computer rooms

Gartner – Over 500,000 in 2010, IDC – under 5,000 in 2008  - over a factor of 100 difference

 

This is mainly skewed by the 400,000+ rack/computer rooms in the Gartner report, so let’s eliminate them since they probably won’t use 100G and see what we get:

 

Number of data centers over 750 sq ft:

Gartner – Over 22,000 in 2010, IDC – under 5,000 in 2008 – still a factor of 4 off

 

Number of data centers over 15,000 sq ft excluding computer rooms

Gartner – 1,110 in 2010, IDC – 2,440 in 2008 – still a factor of 2 off but in the opposite direction

 

A 15,000 sq ft data center holds

Gartner – 500 racks of servers in 2010, IDC – 500 servers in 2008

 

I was hoping to add some diversity to the conversation with more relevant data, but the data muddies the water.  I’ve tried to keep from interpreting the data in any way and just presenting it.  The group might have to decide the best thing to do based on our gut feelings instead of correlated data from analysts, since we don’t have correlation. 

 

Kind regards,

Scott

 

 

 

 

 

From: Alan Flatman [mailto:a_flatman@xxxxxxxxxxxxx]
Sent: Wednesday, December 07, 2011 7:39 AM
To: Scott Kipp; STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Interesting data, Scott. Thanks for sharing. This is a very different profile of data centre sizes compared to the IDC analysis I used in flatman_01_0311 (copy of the IDC data attached). Have things changed so much in 5 years?

 

From the Gartner data you can determine the server population for each data centre size category (I assume we can estimate servers from the number of cabinets - I did this calculation with IDC data in slide 4 of the attached). You can then derive the cumulative length distribution for 100G aggregation links and trend this to 2015 using the Gartner data. I would assume a rectangular footprint to get worst case lengths.

 

Best regards,  Alan

 

----- Original Message -----

From: Scott Kipp

Sent: Tuesday, December 06, 2011 6:24 AM

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Alan,

 

Thanks for the information.  I have found some other relevant information about data centers and the type of links being deployed.

 

First off, Gartner has done a summary and prediction of the number of data centers around the world in their report: Forecast: Data Centers, Worldwide, 2010-2015.  They predict that less than 2% of data centers are larger than 3,000 sq ft and have over 100 racks.  Here’s the summary: 

 

 

Rack Range

Square-Foot Range

2010

 

2010

 

2015

 

 

2015

Rack/Computer Room

2-25

Up to 750

488494

95.6%

 

436441

 

94.8%

Midsize DC

26-100

Up to 3,000

15669

3.07%

 

16013

 

3.48%

Enterprise DC

101-500

Up to 15,000

5668

1.11%

 

6240

 

1.36%

Large DC

500+

15,000+

1110

0.22%

 

1361

 

0.30%

 

They predict that 1,361 data centers will be larger than 15,000 sq ft in 2015.  A 15,000 sq ft data center would likely have the dimensions of 200’ by 75’ or 100’ by 150’.  The major change going on in data centers is that computer rooms are decreasing in numbers while the other sizes are growing. 

 

I also reviewed the Lightcounting data from last year’s Lightcounting Optical Communications Market Forecast and compared LR to SR volumes.  In 2006 – 4 years after the standard was released, about 36% of the 10GBASE-LR and 10GBASE-SR transceivers were LR.  By 2010, the percentage dropped to 20% and it is expected to be 18% in 2011. 

 

Lightcounting expected 100GBASE-LR4 to be about 25-28% of the LR4 and SR10 shipments for several years. 

 

I recently visited a data center that have significant deployments of single-mode fiber in Mainframes over Fibre Channel (FICON).  All FICON deployments use single-mode fiber.  I’ll present on this data center visit in January if possible because it was an interesting case study in how data centers grow organically in odd ways.

 

Kind regards,

Scott

 

From: Alan Flatman [mailto:a_flatman@xxxxxxxxxxxxx]
Sent: Tuesday, November 29, 2011 12:49 AM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Hi folks,

 

Apologies for a late entry into the debate on fibre deployment in the DC. None of the recent Flatman surveys has examined this but I've spoken to BSRIA and DC infrastructure designers. I haven't got all the answers but the following might be useful:

 

BSRIA did a survey of cabling in ~1,000 US data centres of different sizes. This was conducted late 2007 and published in 2008. The attached extracts indicate the presence of OM1, OM2, OM3, OM3+ and SMF. When interpreting this, remember that this is the percentage of data centres where each of these is present - it doesn't estimate the volume in each - it's the "ticks in the box" from 250 respondents who are each responsible for an average of 5 DCs. BSRIA has just (last week) published another survey from DCs in 6 major countries but I haven't seen this yet. Unfortunately, BSRIA do not report on fibre count (i.e. the number of fibre elements in a single cable).

 

Jonathan Jew designs infrastructure for large US DCs. Jonathan's input on SMF deployment is below. Jonathan is copied with this mail.

 

"Alan,

 

I could see 100G Ethernet with a 2 km reach being used not only for mega data centers, but also for:

 

1)     Data centers where the primary and backup data center are within 2 km. 

2)    Data centers where the primary data center and expansion of the primary data center are not contiguous (i.e. the original data center ran out of space and they decided to expand in a nearby but not adjacent location).

3)     Between separated production and development/test data centers where high speed connections are desired to obtain copies of large volume production data for development/testing.

 

That said, the SMF count varies widely between customers with the low for either Mega data centers or data centers falling into one of the categories I’ve listed above at 144 fibers and the high being around 4320 fibers (between data centers within a campus   I am usually involved in the design phase and the initial occupancy of the data center, but typically am not involved in the maintenance of the cabling system afterward.    That said, many of enterprise clients don’t fully utilize their single-mode fiber – the one that installed 144 fibers is using only about 48 fibers now.  The rest were originally allocated for IBM mainframe channels, but they have replaced them with fewer higher speed SAN connections (mostly Fibre Channel).  For my client that installed 4320 fibers, my guess is that they have utilized most of the fibers because they mentioned that they are always running out of fibers for new projects.  That said however, it may be that they will be able to replace multiple lower speed (maybe 10G) connections with an 8 fiber 100G connection and thus reduce their overall requirements.

 

Jonathan"

 

Best regards,  Alan

 

----- Original Message -----

Sent: Monday, November 28, 2011 8:24 PM

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Mike,

You make good observations.  But please note that the ~80/20 split includes all usage not just within a data center.  The barbieri reference cited below also shows on slide 9 that for SONET/SDH the percentage of sub 7km SM optics deployed is less than 20% of the overall SM mix.  And this is for an application that sees much less MM deployment compared to Ethernet.  So while I think your argument has validity, the ratios may need adjustment.  Instead of assuming 20% SM ports in DCs, the SONET/SDH data suggests that it may be more accurate to assume less than 1/5th of this (< 4%) of the total SM ports are in DCs.  So rather than 20% + 0.8*(SM coverage), it may be better to use 4% + 0.96(SM coverage). 

 

Paul

 

 


From: Mike Dudek [mailto:mike.dudek@xxxxxxxxxx]
Sent: Monday, November 28, 2011 1:25 PM
To: Kolesar, Paul; STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: RE: [802.3_100GNGOPTX] Emerging new reach space

 

On the topics of volume I think we need to be clear that the 20% SMF to 80% MMF is not a fixed ratio.    I think that the ratio should be mainly explainable by the capabilities of the units.   At the moment the MMF is sufficiently lower cost than SMF that within the data center most installations use MMF when they can and SMF when MMF hasn’t got the reach.    If the MMF reach becomes shorter it’s volume will drop and the SMF volume will increase.   That is evident in Paul’s spreadsheet (and is I think one of the basic assumptions in his spreadsheet).    Looking at that spreadsheet however there is obviously some additional usage for SMF at 10G because if one puts 300m for the multimode solution it predicts that the SMF percentage is well less than 2%.     Based on this I think we could assume that the SMF percentage will be approximately 20% + .8*(SMF percentage from Paul’s spreadsheet) while the MMF percentage will be .8*(MMF percentage from Paul’s spreadsheet.).     As an example if one assumes the following:

 

MMF reach is dropped to 75m

there are equal numbers of server to switch links using fiber and switch to switch links  

the switch to switch links use a 2:1 ratio for the single versus double-link

 

Paul’s spreadsheet would predict that in his population  there is only a 16.4% single mode usage.    That would change the overall ratio to 33% SMF to 67% MMF.  

 

Adding an additional lower cost shorter distance SMF spec would however complicate the analysis.    The big question would be how much of the 20% additional SMF market would go to the lower cost shorter distance spec.  If a large percentage does then the volume of the LR4 would drop potentially increasing its cost substantially.

 

 

Mike Dudek 

QLogic Corporation

Senior Manager Signal Integrity

26650 Aliso Viejo Parkway

Aliso Viejo  CA 92656

949 389 6269 - office.

Mike.Dudek@xxxxxxxxxx

 

 

From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxxx]
Sent: Thursday, November 24, 2011 6:28 AM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

David,

Thanks for your efforts on this fundamentally important topic, going so far at to involve the governing authority of IEEE for guidance and reassurance of the current practices.  While I still believe that the current practice is overly restrictive, I also appreciate the line that may be crossed if discussions stray to far afield on the topic of price.  I intend to abide by the guidance you have given. 

 

All,

With regard to translation between cost and price, there are several forces at work.  I previously mentioned the gorilla in the room.  There is also the equally important factor of production rate, for increased volumes often translate into decreased costs.  The comparison of cost must be couched in terms of volume, for it can be imbalanced to compare input costs of different proposals if they are stated with expectation of differing volumes.  I urge all contributions on costs to also provide them in the context of their volume assumptions.  We must also realize that if volume context is given in terms like “at equivalent volumes of X”, where X is volume of a baseline technology that is well along its maturity curve, such volumes for 100G technologies may be many years into the future.  Therefore a large gap may exist between present costs and distant future cost potential.  We must also remember to consider the fact that historically the market demand for our optical solutions tends to fall along an 80/20 rule, with 80% of the port volumes going to multimode and 20% going to single-mode. See http://www.ieee802.org/3/hssg/public/nov06/barbieri_01_1106.pdf

 

So we must be weary of using the volumes of multimode solutions as the volume basis for cost projections of proposed single-mode solutions unless those single-mode solutions are presumed to replace multimode solutions and not just augment them in data centers.

 

The above contribution also provides historical adoption rates of single-mode solutions by reach that can set context for expectations.    

 

Regards,

Paul

 

-----Original Message-----
From: Law, David [mailto:dlaw@xxxxxx]
Sent: Thursday, November 24, 2011 1:52 AM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Hi Paul,

 

I want to bring your attention to the text 'Discussion of the cost of inputs necessary to create a compliant implementation of a standard are treated differently from discussions of prices at which compliant implementations can or should be sold. There is no useful or appropriate reason to discuss selling prices of implementations - each implementer of the standard should use its own independent business judgment to make that decision. In contrast, there is a legitimate reason to discuss costs of inputs used in implementation.'.

 

I believe this agrees with your point that the information is publically available - but states 'There is no useful or appropriate reason to discuss selling prices of implementations' which I believe is what you are requesting.

 

If a discussion of current and past end user cost (price) of products that implement the standard were to occur it could very easily stray into a discussion of the future end user costs of the proposed standard, which you agree is not allowed. This is addressed in the answer to the question 'So is it okay to talk about prices or output levels in an IEEE-SA meeting as long as we don’t reach an agreement?' which states:

 

No, it’s not okay. First, you can’t always control where the discussion will go – it may end up in undesired areas. Second, if agreeing on the subject would be unlawful (such as the respective selling prices of compliant products), then that subject should not be discussed. And third, it’s not up to you to decide whether your words and conduct amount to an agreement – in the U.S., that decision gets made by a judge using the peculiar rules of evidence that only courts use and by a jury that is unlikely to know anything about your industry or business. The whole question about your actions will come up after the fact, and with the sure vision of hindsight, any questionable discussion or debate could be seen to have led to a tacit if not an explicit agreement that is prohibited by law. Please do not put the IEEE, your company, your colleagues in the standards community, or yourself at risk by discussing these topics.

 

I have discussed your concerns with David Ringle, Director, Governance & Technical Committee Programs at the IEEE-SA, and he is in agreement with my position.

 

Best regards,

  David

_______________________________________

 

From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxxx]

Sent: 17 November 2011 17:22

To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

David,

Thanks for the pointers to the IEEE policy.  I have been reviewing it and find the following particularly relevant to the thread below.  I have underlined key parts for easier reference. 

 

What else can we discuss? IEEE wants you to have the maximum flexibility to discuss topics relevant to developing a standard while also adhering to certain rules designed to minimize risk. It is impossible to identify all the topics that you can discuss, but here are some that you cannot discuss:

• prices at which products or services implementing the standard should be sold (“price” includes discounts, terms, and other conditions of sale);

• profits or profit margins;

• individual companies’ market shares or sales territories;

• allocation of customers, markets, production levels, or territories; or restricting the customers to whom, or territories in which, a company may sell or resell products;

• using standards or certification programs to exclude suppliers or competitors from the marketplace for any reason other than cost-performance or technical considerations;

• conditioning the implementation of a standard on the implementer’s use of products or services from a particular supplier [such as requiring use of a particular manufacturer’s components or requiring implementers to use a particular service provider(s) for compliance certification];

• bidding (or terms of bids) or refraining from bidding to sell any product or service;

• any matter which restricts any company’s independence in setting prices, establishing production and sales levels, choosing the markets in which it operates, or the manner in which it selects its customers and suppliers.

 

The first bullet gets at the heart of the matter.  It is what everyone keys on when the discussions of price are thwarted.  It states we cannot discuss what prices should be.  I think we can all agree that to do so would be anti-competitive.  But it says nothing about discussions of prices as they actually are in the market today or have been in the past.  To muzzle such discussions is overstepping the bounds and scope of the policy, for there is no anti-trust reason why prices that are in public domain, such as can be found posted on public web sites, cannot be brought forward.  Such activity is not price fixing, it is price reporting.

 

In the case of setting new standards we are naturally talking about documents that affect the future.  This is the context under which the IEEE policy is written and should be interpreted and applied.  When the policy says:

 

Discussion of the cost of inputs necessary to create a compliant implementation of a standard are treated differently from discussions of prices at which compliant implementations can or should be sold. There is no useful or appropriate reason to discuss selling prices of implementations – each implementer of the standard should use its own independent business judgment to make that decision. In contrast, there is a legitimate reason to discuss costs of inputs used in implementation.

 …

 

With regard to the costs of inputs used in implementing a standard, the only permitted discussion is the degree to which such costs may differ. Examples of permissible discussion topics would include differences in comparative component costs, operating costs, licensing costs, or the aggregate of such costs. The importance of this restriction on discussion is reinforced by the understanding that participants in the development of a standard often come from multiple stages of the supply chain (e.g., the input cost of a component to a system manufacturer is the output price of a component supplier).

 

Thus, in standards development technical activities, participants may discuss the relative costs (in terms, for example, of percentage increases or decreases) of different proposed technical approaches in comparison with the relative technical performance increases or decreases of those proposals.

 

It is obviously referring to the setting of new standards.  This is conveyed by the use of the word “proposed” and more fundamentally by the entire context of the policy for guiding standards development.  While it is not referring to discussions of cost for implementations already standardized and in the market, the above restrictions are needlessly being applied that way.  There is no anti-trust implication to discussions of publicly available information, even if it is with regard to price.  How can there be?  It’s already fully available to all participants. Such restrictions would imply that discussions of current prices somehow set agreements on future prices of yet-to-be-standardized implementations.  That is a non-sequitur. 

 

If we are talking about input costs on items affecting the setting of future standards, the guidelines are clear that we must use relative costs, not absolute costs.

 

It also recognizes the supply chain issue, that one link’s cost is the previous link’s price.  But it places no restriction on the point in the supply chain at which relative costs can be compared.  Therefore imposition of restrictions on discussions of relative costs at the top of the supply chain or any other specific part of the supply chain is outside of this policy.

 

While I do not relish raising these issues, I feel compelled to do so because after reading the actual policy it is clear to me that our practices in 802.3 are unnecessarily restrictive both to the letter and spirit of the IEEE policy.

 

Regards,

Paul

 

-----Original Message-----

From: Law, David [mailto:dlaw@xxxxxx]

Sent: Thursday, November 17, 2011 8:02 AM

To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Hi Jonathan,

 

While I note that Dan has already stated that comparing cost at the end user is not permitted, and discussion did not go into that area, I wanted to reiterate that such discussion are not permitted. I also wanted to respond to your request to me to comment.

 

As I have often stated at meetings, I would encourage everybody to read the IEEE policy document 'Promoting Competition and Innovation: What You Need to Know about the IEEE Standards Association’s Antitrust and Competition Policy' <http://standards.ieee.org/develop/policies/antitrust.pdf>. I personally think it provides a good explanation of why these rules are in place, due to the Antitrust and competition laws throughout the world, and the consequences if they are not followed.

 

This policy states that we cannot discuss, among other items, 'prices at which products or services implementing the standard should be sold (“price” includes discounts, terms, and other conditions of sale)'. This is enforced in subclause 5.3.10.3 'Discussion of relative cost/benefit analyses' of the IEEE-SA Standards Board Operations Manual and stated in the 'Guidelines for IEEE-SA Meetings' slide shown at every meeting.

 

By way of an introduction to why these rules are in place, and the consequences to all of us if they are not followed, I offer the following extract from the introduction to the above policy document. Clause 2 of this document addresses 'Cost Discussions' in further detail. Participants may also wish to consult 'What You Need to Know About IEEE Standards and the Law' <http://standards.ieee.org/develop/policies/stdslaw.pdf>.

 

----

 

Promoting Competition and Innovation: What You Need to Know about the IEEE Standards Association’s Antitrust and Competition Policy

 

Antitrust and competition laws throughout the world rest on the premise that competition in the provision of products and services is the best way to ensure that consumers and other users receive maximum innovation and quality at the lowest possible prices. But sometimes effective competition requires a measure of cooperation among competing firms.

 

Standards development is one of those areas. Standards development serves one part of the IEEE’s mission – advancement of global prosperity by fostering technological innovation – but it can do so only if the standards development is conducted consistent with the antitrust and competition laws that regulate the nature and extent of cooperation in which competitors can legitimately engage.

 

The IEEE-SA is an international membership organization that provides a standards program serving the global needs of industry, government, and the public. A violation (or claims of violation) of competition laws will jeopardize what all participants are working so hard to build; will impede the IEEE mission; and may expose participants and their employers to the risk of imprisonment and other criminal penalties, civil remedies, and significant litigation costs. Even if a competition-law case or investigation is ultimately dropped, that will often happen only after the parties have spent considerable resources in responding to information requests and defending against the claims.

 

The IEEE-SA wants to help all of its participants avoid competition-law problems. Many IEEE-SA participants receive antitrust/competition-law compliance training from their employers, and IEEE-SA participants should always consult with their own or their company counsel when they have competition-law-related questions. This brochure is not intended to replace that competition-law training, advice, or other competition-law-related resources that participants may have available to them; rather, this brochure is intended to highlight the competition-law risks that are most pertinent to standards development and to explain the IEEE-SA’s policies with respect to competition law matters.

 

<http://standards.ieee.org/develop/policies/antitrust.pdf>

 

----

 

Best regards,

  David

 

________________________________________

From: Daniel Dove [mailto:ddove@xxxxxxx]

Sent: 16 November 2011 20:04

To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Hi Robert,

 

I am not sure its even appropriate to comment about prices-of-resale. Its a topic better left understood without discussion within the IEEE. We all have a general idea of the challenges we face in assessing relative cost.

 

In my opinion, you do it by looking at fundamental architectural components and their specifications. You do it by looking at cost from a form/fit/function perspective and leave out things like market pricing, demand or volume. You choose *cost* comparisons with components that have equivalent markets/volumes and then your comparison can have merit without getting into the esoteric aspects of the market.

 

Dan

 

From: Robert Lingle <rlingle@xxxxxxxxxxxxx>

Reply-To: Robert Lingle <rlingle@xxxxxxxxxxxxx>

Date: Wed, 16 Nov 2011 22:52:26 +0000

To: 100G Group <STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx>

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Jonathon,

 

I agree with your former sentiment about it being against IEEE guidelines, so we must not do it. I am sure Paul agrees.

 

However I don’t agree with the latter sentiment. I think Paul is pointing out a real quandary in our task. I think we can point out cases where the adoption of components has been substantially impacted by the prices-of-resale, over a lengthy period of time. I am not sure how long we would have to wait for the long term?

 

It seems to be a genuine difficulty in carrying out the analysis we all want to see.

 

I don’t have a solution.

 

Warm regards,

Robert

 

From: Jonathan King [mailto:jonathan.king@xxxxxxxxxxx]

Sent: Wednesday, November 16, 2011 5:15 PM

To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Hi Paul,

 I cannot support comparing  cost at the end user, because (in my opinion) that is quite clearly a price, and would be contrary to IEEE guidelines.

 I would like David Law to weigh in on this.

 

In an efficient market economy, the cost of manufacture (not the cost of distribution) of a product, or group of products, is the best indicator for long term cost to the end user.  I think that’s what we (802.3) want to get some indicator of when striving for  ‘low cost’ solutions.

Best wishes

Jonathan

 

 

From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxxx]

Sent: Tuesday, November 15, 2011 8:30 PM

To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

I also like Jack’s perspective foundation.  And I have to agree that a common small port form factor would be valuable.  

 

Regarding the PMD cost comparisons, I also agree that picking a common cost basis could eliminate translation problems that come with use of different bases.  For PMD relative costs I am fine with choosing either 10GE SR or LR because I don’t quite appreciate Chris’ concerns over the danger in using LR as the basis.  But if there is a problem, it seems like it could be avoided by eliminating the bit/sec factor and instead using simple cost.  

 

However, there is still the issue of picking the point in the supply chain at which the baseline cost is chosen.  It could be at the level of the transceiver manufacturer, or up one level to the switch manufacturer, or off on another branch to distributors, or up one level again to the end user.  There can be substantial differences in these costs, which is what likely gives rise to the “order of magnitude” perception disparity that Chris mentioned.  While we are forbidden to talk about it, it is the 500 pound gorilla in the room.  In some cases it tips the scales lightly and in other cases it leans its full weight.  Yet this weight is what is creating a 100GE market acceptance barrier in data center deployments.  

 

While there is clear cost reduction potential in reducing the multimode lane count based just on the improvement in media cost, the benefit of defining a new single-mode solution may depend largely on the gorilla, for if we go to a parallel solution the media will only get more costly.  So for me the biggest cost question is whether defining a new single-mode solution moves the gorilla from leaning heavily to leaning lightly. 

 

While I’d like to agree to use cost at the transceiver maker level as the basis, it cannot address this big question.  If used it will likely result in highly distorted perceptions and projections because it ignores the order-of-magnitude problem.  The only point in the supply chain where transceiver costs and cabling costs come together is at the end user level.  This dilemma is hitting me squarely as I try to further my work on the “Kolesar_Kalculator”.   I continue to wonder what good this tool will do if the PMD cost and cable cost are in two different silos without a common basis to bring them together, for it is the combination of the two that defines the total channel cost that we must optimize.  I need to find a way to appease the gorilla.  If only Fay Wray were here…   

 

________________________________________

From: Chris Cole [mailto:chris.cole@xxxxxxxxxxx]

Sent: Tuesday, November 15, 2011 7:33 PM

To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx

Subject: Re: [802.3_100GNGOPTX] Emerging new reach space

 

Hello Jack,

 

Nice historical perspective on the new reach space.

 

Do I interpret your email as proposing to call the new 150m to 1000m standard 100GE-MR4? ☺

 

One of the problems in using today’s 100GE-LR4 cost as a comparison metric for new optics is that there is at least an order of magnitude variation in the perception of what that cost is. Given such a wide disparity in perception, 25% can either be impressive or inadequate.

 

What I had proposed as reference baselines for making comparisons is 10GE-SR (VCSEL based TX), 10GE-LR (DFB laser based TX) and 10GE-ER (EML based TX) bit/sec cost. This not only allows us to make objective relative comparisons but also to decide if the technology is suitable for wide spread adoption by using rules of thumb like 10x the  bandwidth (i.e. 100G) at 4x the cost (i.e. 40% of 10GE-nR cost) at similar high volumes.

 

Using these reference baselines, in order for the new reach space optics to be compelling, they must have a cost structure that is referenced to a fraction of 10GE-SR (VCSEL based) cost, NOT referenced to a fraction of 10GE-LR (DFB laser based) cost. Otherwise, the argument can be made that 100GE-LR4 will get to a fraction of 10GE-LR cost, at similar volumes, so why propose something new.

 

Chris

 

From: Jack Jewell [mailto:jack@xxxxxxxxxxxxxx]

Sent: Tuesday, November 15, 2011 3:06 PM

To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx

Subject: [802.3_100GNGOPTX] Emerging new reach space

 

Following last week's meetings, I think the following is relevant to frame our discussions of satisfying data center needs for low-cost low-power interconnections over reaches in the roughly 150-1000m range. This is a "30,000ft view,"without getting overly specific. 

Throughout GbE, 10GbE, 100GbE and into our discussions of 100GbE NextGenOptics, there have been 3 distinct spaces, with solutions optimized for each: Copper, MMF, and SMF. With increasing data rates, both copper and MMF specs focused on maintaining minimal cost, and their reach lengths decreased. E.g. MMF reach was up to 550m in GbE, then 300m in 10GbE (even shorter reach defined outside of IEEE), then 100-150m in 100GbE. MMF reach for 100GbE NextGenOptics will be even shorter unless electronics like EQ or FEC are included. Concurrently, MMF solutions have become attractive over copper at shorter and shorter distances. Both copper and MMF spaces have "literally" shrunk. In contrast, SMF solutions have maintained a 10km reach (not worrying about the initial 5km spec in GbE, or 40km solutions). To maintain the 10km reach, SMF solutions evolved from FP lasers, to DFB lasers, to WDM with cooled DFB lasers. The 10km solutions increasingly resemble longer-haul telecom solutions.!

  There is an increasing cost disparity between MMF and SMF solutions. This is an observation, not a questioning of the reasons behind these trends. The increasing cost disparity between MMF and SMF solutions is accompanied by rapidly-growing data center needs for links longer than MMF can accommodate, at costs less than 10km SMF can accommodate. This has the appearance of the emergence of a new "reach space," which warrants its own optimized solution. The emergence of the new reach space is the crux of this discussion.

Last week, a straw poll showed heavy support for "a PMD supporting a 500m reach at 25% the cost of 100GBASE-LR4" (heavily favored over targets of 75% or 50% the cost of 100GBASE-LR4). By heavily favoring the most aggressive low-cost target, this vote further supports the need for an "optimized solution" for this reach space. By "optimized solution" I mean one which is free from constraints, e.g. interoperability with other solutions. Though interoperability is desirable, an interoperable solution is unlikely to achieve the cost target. In the 3 reach spaces discussed so far, there is NO interoperability between copper/MMF, MMF/SMF, or copper/SMF. Copper, MMF and SMF are optimized solutions. It will likely take an optimized solution to satisfy this "mid-reach" space at the desired costs. To repeat: This has the appearance of the emergence of a new "reach space," which warrants its own optimized solution. Since the reach target lies between "short reach" and "long reach," "mi!

 d reach" is a reasonable term

Without discussing specific technical solutions, it is noteworthy that all 4 technical presentations last week for this "mid-reach" space involved parallel SMF, which would not interoperate with either 100GBASE-LR4, MMF, or copper. They would be optimized solutions, and interest in their further work received the highest support in straw polls. Given the high-density environment of datacenters, a solution for the mid-reach space would have most impact if its operating power was sufficiently low to be implemented in a form factor compatible with MMF and copper sockets.

Cheers, Jack