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

Re: [802.3_100GNGOPTX] Agenda and Presentations



Hi Brad,

 

Let’s agree that we have different thresholds for what is similar enough to be contained in one objective. I am comfortable having a single objective resulting in specifications that support Cat 5 and Cat5E, or OM3 and OM4. In contrast, I would not be comfortable with a single objective for Cat 5 (or Cat 5e) and Cat 6A, or OM1 (or OM2) and OM3. The mental test I apply is to consider how an end user, not present in the room when the decision was discussed and reached, react when he or she heard the result.


With respect to 802.3ba and parallel, we did have to face the problem of how do we handle the desire to have both parallel and duplex MMF for 40GE. We decided that 40GE duplex MMF would require a separate objective from parallel 40GE MMF objective, and then dropped it because yet one more objective might have been the straw that broke the .ba camel back. So if we are looking for clear precedent, .ba made a distinction between parallel and duplex.

 

Chris

 

From: Brad Booth [mailto:Brad_Booth@xxxxxxxx]
Sent: Sunday, January 22, 2012 8:47 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Agenda and Presentations

 

Chris,

My apologies. I interpreted your statements as an implication that some may try to perform a bait-and-switch.

If the process is followed, then I agree that any modifications to the objectives by the task force shall ensure that the other documentation (5 criteria and PAR) continue to be relevant.

As for your 1000BASE-T example, you may wish to review the actual cable specified by 802.3ab vs. what you list. The Task Force dealt with channel models. Considering Cat 5 was not tested with all the losses necessary for a 1000BASE-T system, the Task Force added other required parameters necessary for 1000BASE-T operation. And so, Cat 5e was "born".

In many cases, the specification for cabling coincides with the start date for a .3 standard. OM3, Cat5e, Cat6A and OM4 were not found in previous generations of Ethernet standards, neither was the concept of parallel fibers until .3ba.

As you said, it will be up to the study group and ultimately the working group to decide if the objectives are viable for a project. The question of parallel vs. a PAM WDM solution for 100G SMF is certainly going to be an interesting one, but it won't be the first we've broken new ground. :-)

Cheers,
Brad



-----Original Message-----
From: Chris Cole [chris.cole@xxxxxxxxxxx]
Sent: Friday, January 20, 2012 04:01 PM Central Standard Time
To: Booth, Brad; STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: RE: [802.3_100GNGOPTX] Agenda and Presentations

Hi Brad,

 

Thank you for your comments.

 

You made an error of logic in concluding that my point was that “bait-and-switch” tactic was used. My point was that changing an implicit cabling objective in Task Force will result in many in the larger optics community feeling misled. Basic causality theory states that if a specific tactic causes a specific result,  that does NOT imply that observing the specific result means it was caused by the specific tactic. There could be many other tactics that could have caused the observed result.

 

Further, “bait-and-switch” tactic requires pre-meditation which gets into motivation. In general, I stay away from attributing motivation because it is very hard to objectively determine what is in the minds of others. In this case I will  make an exception since speculation about motivation has been brought up, and will speculate as to the motivation of those promoting the multiple applications under one objective umbrella. My suspicion is that they would like to see us get through the 5 Criteria Gate soon, so that we can move to Task Force and they can present a detailed case to persuade others to adopt their technical solution to meet the application they view as important. Nothing terribly sinister about this.

 

I fully agree with you that we should follow 802.3 precedent to leave the selection of the best solution up to the Task Force and objectives should not constrain this. However this precedent is used for picking a solution to the transmitter and receiver, and not to cabling. The 802.3 precedent for cabling is just the opposite, specifically to tightly specify the type of cable used in the five criteria objectives. To my knowledge every 802.3 project, other than 802.3ba has completely restricted the cable choice so that if the Task Force wanted to change the cable type,  it would have to go back  and change its objectives.

 

This has been the case for every copper cable project.

 

Let’s look at the Compatibility Criteria Response for 802.3ab (1000BASE-T):

 

“In a fashion similar to the 100BASE-T standard, the current physical layers will be replaced with new Physical Layers (PHY) as appropriate for 1000 Mb/s operation over the defined link: a link which meets the  link requirements of 4 pair Category 5 100 ohm balanced copper component specifications {…} 1000BASE-T offers the maximum compatibility with the current installed base of more than 60 million CSMA/CD nodes, most of which {are} copper cabling systems.”

 

There is no doubt here about what type of cable to will be used, and definitely it is NOT left up to the Task Force to decide.

 

This has also been the case for almost every optics project, which has always restricted the type of cabling to be used, it is either duplex MMF or duplex SMF. Definitely it has never been left up to the Task Force to decide what type of fiber cable to use, MMF or SMF. Some optics projects, for example LRM that you cite,  constraint the cabling choice even tighter. Here is an excerpt from the Broad Market Potential Criteria response for 802.aq (10GE-LRM):

 

“Clearly there is a need for a low cost solution that will utilize the installed base multimode fiber infrastructure. Interest in 10Gb/s on FDDI-grade multimode fiber has been demonstrated by the attendance of more than 156 vendor and user representatives at CFI meeting … “

 

The following excerpt from the 802.3aq Technical Feasibility Criteria Response makes this even clearer:

 

“Presentations made to 10Gb/s on FDDI grade multimode fiber Study Group illustrate the technical feasibility of 10Gb/s signaling using structured fiber cabling as defined by IEC 60792-2.”

 

No doubt here that the Task Force has no discretion to change the cable type. If the Task Force decided to change to supporting anything other than OM1/OM2 duplex MMF, it would have had to go back and get its objectives changed, which would have required Working Group approval. So not only do we constrain the cabling fiber diameter (MMF vs. SMF), but we also constrain the properties of that fiber. This is exactly the same as in all the copper projects.

 

The one exception has been 802.3ba, where parallel fiber came into consideration, and we did not differentiate between duplex and parallel. However, as Steve Trowbridge correctly pointed out, that was an oversight we could get away with because there was unanimous implicit agreement to parallel cabling for MMF and duplex cabling for SMF. Anyone that came in and proposed otherwise would have been lynched.

 

If we look at 802.3 optic objectives precedence for cabling, it can be argued that the difference between MMF and SMF duplex fiber, is less than the difference between duplex MMF and parallel MMF cabling, or duplex SMF and parallel SMF cabling. Again, using the LRM example you  brought up, the very first LRM modules that were developed by myself, Sudeep, Jonathan, and our colleagues were X2 modules that supported both 10GE-LR 10km SMF specifications and 10GE-LRM 220m OM1/OM2 specifications. The first 22 of these modules were delivered to none other than our chair; Mr. Dan Dove. So optical modules can be built at a given aggregate data rate to support both MMF and SMF cabling. However no specifications or modules can be  built that support both parallel and duplex cabling at the same aggregate bit rate.

 

So we can chose to adopt NG 100G OE SG 5 criteria objectives that do not specify the cabling type, but that will be a dramatic departure from all past 802.3 projects, and the decision should not be taken lightly, and has to involve the full Working Group in deciding that we will starting doing things very differently from how 802.3 has always done them in the past.


Chris

 

From: Brad Booth [mailto:Brad_Booth@xxxxxxxx]
Sent: Friday, January 20, 2012 1:34 AM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Agenda and Presentations

 

Chris,

 

I am concerned about your statements.

 

While I do understand the point you’re making about bait-and-switch, it is not a good idea to overly presume a solution when setting the objectives and documenting the 5 criteria.

 

I’m sure that you’d agree that the objectives and criteria should be written such that a broad range of possible solutions can be presented to the task force (once formed), and for the task force to decide the solution that they feel best meets the objectives and the criteria. It would also be preferential to not create multiple solutions to satisfy the shorter SMF demand. Per all the information shown today, this is a small market and two solutions would be ineffective in satisfying broad market potential. Let’s learn from 10G and try not to create LX4 and LRM solutions that no one uses today.

 

As a further reminder that the SG should not be pre-supposing a solution: the final 10GBASE-T solution that was selected by the task force was far different than what was under consideration during the study group phase of the project, but the PAR, 5 criteria and objectives were broad enough to enable the task force to make what they felt was the best decision.

 

Cheers,

Brad

 

 

From: Chris Cole [mailto:chris.cole@xxxxxxxxxxx]
Sent: Thursday, January 19, 2012 7:06 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Agenda and Presentations

 

Hi Dan,

 

I would like expend on your comment that we have a jam-packed agenda. Not only do we have a large quantify of presentations, but we also have great depth and breadth of material in those presentations. Some of the technical results being presented are ground breaking and reflect major efforts. We should take pride in our Study Group for being a venue for reporting such high caliber work.


However, this does lead me to several concerns. The first concern is that we are rushing into adopting an SMF objective without sufficient justification.

 

Let’s take as an example Sudeep’s excellent presentation; bhoja_01_0112. As pointed out by Mark in nowell_01_1111 (page 8 table), PAM-16 has 1 laser instead of four lasers required for LR4. All things being equal that can translate to dramatic long term cost reduction, which Gary reconfirms in nicholl_01_0112. One seductive analogy that comes to mind is 10GE-LR versus 10GE-LX4. However, it is the implementation details that will determine whether this promise is real or an illusion. The problem is that many Study Group participants, including us, have no experience with high order PAM modulation, to enable a quick conclusion. We will need lots more time to digest this material.

 

A great example of why this understanding is critical comes from nowell_01_1111 observing that xA-yDPSK modulation also requires 1 laser instead of 4 lasers. There are published works showing Si implementations, so even the building blocks appear similar to the PAM-16 proposal. Fortunately, many Study Group participants already have multi-year experience with complex modulation and can immediately conclude that this approach is orders of magnitude away from being technically feasible as a low cost, low power client interface.

 

If we want to seriously consider PAM-16, we will need time to study it and understand the real implementation limitations. Mark and Gary have convinced me that this an area worthy of further study, but I do not see how that can be done by next week as urged by Dan. We will need at least two more meeting cycles to reach an independent conclusion if this is real. It is critical for the chair to have a sense of urgency to make forward progress. I fully internalized this during 802.3ba, where without John D’Ambrosia continuously herding the participating cats, it would have taken us a decade to finish.  However, that needs to be balanced against allowing sufficient time for proper deliberation. if we want a SMF objective, we have to invest more time, unfortunately with no assurance that we will end up with an SMF objective at the end.

 

My second concern is that we are lumping together two very different applications and technologies into one SMF objective bucket, specifically parallel SMF for 300m to 500m applications and duplex SMF for 1km to 2km applications. A group of us discussed this at lunch during OIF, and Steve Trowbridge made a very insightful observations. In 802.3ba we did not have explicit statements about fiber cabling in either the MMF or SMF objectives, but very strong implications were there. Every participant understood that the MMF objective was for parallel MMF, and the SMF objective was for duplex SMF. Had we moved into Task Force and switched to duplex MMF (let’s say using WDM) or parallel SMF (let’s say because of cost), many in the larger optics community would have felt misled.

 

Looking at parallel SMF, we are reaching a critical mass for Technical Feasibility, among other reasons because there are multiple credible technical solutions on the table. On the other hand, the Broad Market Potential for 300m to 500m parallel SMF standard is questionable. In additional to carving out a small fraction out of a moderate Market, there are many end users that will not even consider parallel SMF for future deployments. This runs counter to solving fiber congestion which is becoming a major problem in IDCs.

 

For duplex SMF, if we assume the potential for dramatic cost reduction versus LR4, like 4x, then Broad Market Potential is more credible. However, we are nowhere near showing Technical Feasibility.

 

Given this disparity, if we are serious about parallel SMF and duplex SMF, each addressing different applications, we need to two separate SMF objectives, where each one stands on its own merit. To move forward with one SMF objective, by using parallel SMF for Technical Feasibility and duplex SMF for Broad Market Potential would be misleading the larger optics community. Moving forward with one SMF objective by using parallel SMF for both Technical Feasibility and Broad Market Potential, and then switching to a duplex SMF solution would be even worse.

 

I look forward to a great meeting next week and wish everyone safe travels.


Thank you

 

Chris

 

From: Daniel Dove [mailto:ddove@xxxxxxx]
Sent: Thursday, January 19, 2012 7:55 AM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: [802.3_100GNGOPTX] Agenda and Presentations

 

Dear Study Group Members,

 

We have received all of the presentations to be given in Newport Beach and the leadership team is finalizing the agenda shortly.

 

Our vice-chair Kapil Shrikhande will be sending out an email to the reflector later today communicating the link when the website is ready for operation. Our goal is 12pm PST today.

 

As you will see, we have a jam-packed agenda and are planning to work from 8am to 6pm both days.

 

Due to the magnitude of the workload, I will be putting some rules on discussion to ensure fairness and efficiency. We will discuss these at the beginning of the meeting. Your contribution to an efficient flow of work will help us to achieve success.

 

If we can conclude our SG work in Newport Beach, we may find ourselves with less to do in Waikoloa..think about that! :)

 

Best Regards,

 

Dan Dove

Chair, Next Generation 100Gb/s Optical Study Group