| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Hi John,
Glad to hear this discussion as its probably something that needs to be talked about and pondered carefully.
Comments below in red.
On 7/14/13 4:56 AM, John D'Ambrosia wrote:
We had a number of alternatives proposed and none of them (so far) have demonstrated that they can achieve a level of cost, power, size reduction relative to LR4 with a CAUI-4 interface to justify adding a new PMD. We still have a meeting to see if we can come to consensus. We have pored over the challenges of each proposal and are yet to demonstrate all of the 5C are met with them.Interesting thread to follow –
Actually – I have to support Brad regarding the statements of the project documentation needing updating. It is not just about meeting objectives – it is meeting objectives while meeting all criteria responses.
First, Item #5 of BMP states
100 Gb/s Ethernet optical PHY types utilizing a 4 x 25 Gb/s electrical interface, and optimized SMF interfaces will reduce cost, size and power for links in the growing Data Center market and provide a balance in cost between network equipment and attached stations.
IMHO – it isn’t that alternatives haven’t been proposed – it is that no solution has achieved consensus. And I do not believe that the current –LR4 solution with a x4 interface will even be close to a solution for the data center until probably the end of the decade.
I tend to agree that LR4 will be challenged to solve data center fabric needs even with CAUI-4 and packaged in a smaller form factor, but there are some who believe that the cost/power/size reductions available will drive volumes up and create a spiraling reduction in cost. This is to be seen.
The statement is "100 Gb/s Ethernet optical PHY types utilizing a 4 x 25 Gb/s electrical interface, and optimized SMF interfaces will reduce cost, size and power for links in the growing Data Center market and provide a balance in cost between network equipment and attached stations."I will also note that the statement is clear – an optimized SMF interface. If nothing is approved and the TF goes forward without doing an SMF interface. Perhaps the choice of word of “interface” over PHY was specific, however, it is not clear, given the mention of the 4x25 interface prior to the optimized sMF interface. I agree with Brad, the document without “an optimized SMF interface” does not match its 5C statements in their entirety.
Note: We do not say NEW with regard to PHY types and "optimized SMF interfaces" does not mean NEW either. It means optimized. We have heard proposals showing new technology innovations that are going to optimize the cost/power/size of LR4 interfaces to the point that it makes a much more challenging proposition to get below it in cost/power/size. Its still possible, but the language is accurate as written.
Now Robert’s comment is interesting because he points to a carrier and comments about client optics. If you want to say these links exist in data center fine- but the data presented between the .3ba and .3bm TF’s were not just talking about 100G for those connections, they were talking about in their networks as well. So in my opinion Robert’s statement would support a BMP statement focused on the “growing client optics market.” This would be a change to the 5C.
Next, let’s look at the statement Dan pointed to which is #6 under Distinct Identity –
6. The amendment will enable new PHY types over SMF which consist of the existing 100GBASELR4 and 100GBASE-ER4 optical PMDs with The amendment will define a new 100 Gb/s SMF PMD in addition to these if it can be shown that a SMF PMD with a shorter reach than 100GBASE-LR4 has sufficient cost, density, or power difference to justify an additional SMF PMD type.
I appreciate the input. If anything, I think the word "types" could have been written "instantiations". That addresses the point being made and minimizes the changes required. Still, I will think about your suggestion further.
The problem with this statement is that no new PHY type is being created by changing CAUI from 10 lanes to 4 lanes. CAUI is an optional physical instantiation. A new CAUI would only be another option – not another PHY type.
Something that would be better is -
The amendment will define a new 100 Gb/s SMF PMD in addition to these if it can be shown that a SMF PMD with a shorter reach than 100GBASE-LR4 has sufficient cost, density, or power difference to justify an additional SMF PMD type. The amendment will create a new optional physical instantiation, based on four electrical interconnect lanes in each direction, which will enable implementations that help to reduce cost and power, while improving density for implementations of existing SMF PHY types or based on the new 100 Gb/s SMF PMD.
These changes to the 5C responses, IMHO, would bring the document in sync with the 5C responses, which I could support.
My representation of LR4 is that 802.3bm enables lower cost/power/size and leave it to data center operators to determine whether or not it solves their needs. I'm disappointed as well that we have not come to consensus on any of the proposed SMF interfaces, but we have a meeting to go and so will continue to push for a breakthrough in consensus building.I am still disappointed that consensus on a solution to address 100G inside data center networks (not client optics) has not been reached yet. However, I would be further disappointed if the current solution went out being represented as the solution for the data center need.
Brad,
I agree that we need consensus on the question of whether LR4 with CAUI-4 meets our 5C and objectives as well. The Study Group identified LR4 as capable of meeting that objective if no other PMD could achieve sufficient cost/power/size differential, but as a Task Force we will need to confirm that view.
Dan
John
From: Brad Booth [mailto:bbooth@xxxxxxxx]
Sent: Saturday, July 13, 2013 10:54 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] Agenda & Schedule Updated
Dan,
Correct. My statement was based on an opinion. An opinion of someone who has been through this process more than once. :-)
If the LR4 and CAUI4 satisfies the objective, then the task force will need to make that decision with a 75% consensus. In my opinion, any proposal and that includes LR4 with CAUI4 requires 75% approval. It would be very dangerous to simply state that the task force didn't pick an alternative; therefore, LR4 with CAUI4 is the winner. As part of the process, the task force has to agree with a 75% majority that LR4 with CAUI4 satisfies the objective.
Thanks,
Brad
On Sat, Jul 13, 2013 at 3:47 PM, Daniel Dove <ddove@xxxxxxx> wrote:
Hi Chris,
As Brad said, the agenda is subject to review & approval by the Task Force.
The rules I chose to prioritize were the most fair I could come up with as it gave proponents the opportunity to build their own supporter's list, and also it created an incentive to build consensus.
We are going to be pretty challenged for time as our room is going to be shut down at 7p. Once I land in Geneva, I will start working on additional space for continuing our meetings later than 7p.
Brad, I believe your assertion regarding whether or not our project has met its objectives without an additional SMF PMD is based on an opinion. The Study Group clearly stated in our PAR/5C documents that LR4 with a CAUI-4 electrical interface would lower the cost, power and size and thereby potentially meet our objective. The language clearly stated that an additional PMD would only come if it provided sufficient differential in cost/power/density to be justified by the Task Force. So far, the Task Force has not deemed any alternative to meet the necessary criteria.
Perhaps this meeting?
To be successful in adopting an SMF baseline proposal, those who are promoting new PMD approaches should be working to build consensus, and alleviate concerns of those who are not taking a position at this time.
All,
--
Dan Dove
Principal Engineer
Dove Networking Solutions
530-906-3683 - Mobile