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

Re: [HSSG] Soliciting Support for my Presentation

Thanks for elevating the level of debate and I will do the same. 
Regarding economic feasibility of 40G transceivers, I have talked to several transceiver vendors and most predict a near linear growth in cost between 40GE and 100GE at similar volumes.  Thus 40GE transceivers would cost about 40% of the cost of 100GE transceivers.  Others have proposed that there is about a 2X difference in cost between 100GE and 40GE.  Some representatives have acted like a 2X cost difference is not substantial, but I see it as significant. 
These costs do depend on volumes and economies of scale, so predicting volumes in 5 years is a challenge. 
Another aspect of economic feasibility that has not been discusssed much in the HSSG is power consumption.  Power consumption has become one of the leading design constraints and cost points in choosing where data centers are located.  For example, Google and Microsoft are locating their plants near hydroelectric power plants along the Columbia river to save on power costs.  Read more at:
I've heard that 40G transceivers would consume about 40% of the power of 100G transceivers.  Transceiver power consumption adds up and consuming 60% less power in a component is significant.  I would welcome more input from the transceiver vendors about the expected power consumption and initial cost of 40G vs. 100G transceivers. 
Regarding tieing alternative technologies together, 40GE would be a subset of 100GE from my perspective.  It is unfortunate that the 40GE proponents came to the meetings fairly late in the HSSG PAR approval process.  I understand the groups frustration of almost being complete and then a new requirement being suggested.  The intent of the 40GE proponents was never to delay the PAR and we feel that the PAR could progress rapidly with a simple inclusion of 40GE.  Tying alternate proposals together does not make sense unless one is a subset and requires little additional work. 
If a new 40GE PAR was split out and accepted, wouldn't the 40GE group quickly catch up and  possibly wait for the 100GE standard since it plans to piggyback on almost every aspect of the 100GE standard?  I know others have decades of experience in developing IEEE standards.  If some scenarios could be shown as to how the parallel standards might progress, I would be more comfortable with taking this approach.   
It seems like most people agree that 100GE and 40GE are both needed.  The conflict arises in how we get there.
The distinct identity between 100GE and 40GE will probably be the difference between 1 MPO ribbon connector and two.  Only if a 100GE transceiver uses bidirectional links in a single ribbon would they both have a single ribbon interface.  To your point, the QSFP transceiver supplies 4 lanes that could be tied together by a 40GE MAC or four 10GE MACs.  The physical interface could be broken out to individual links with a fanout cable.   A 10X10 link would probably not be broken out with a complex patchcord into ten 10 GE links.  Being able to use the QSFP in two applications may drive volume that decreases price but it may cause the customer to consider what they are cabling.
Regarding clustering PMDs, I guess you're referring to linking four  transceivers (like 10GE SFP+s )  together within the box to form an external 40GE link.  This approach could have some skew problems that would probably be eliminated with the QSFP.   You asked about my thoughts on having OTN on the backplane and I don't have any and did not propose this.  This also falls out of the realm of 40GE being a subset of 100GE, so I would not be in favor of it. 
You asked how do we ensure that 10km and 40 km 40GE PMDs are not introduced. The answer is that companies should not qualify or support non-standard parts.  I don't know of any enforcement agency for standards, so it is up to individual companies.
I look forward to hearing your proposals to move forward.  I'm sorry if I'm belaboring these points but wanted to answer your questions. 

From: Dove, Dan []
Sent: Tuesday, June 12, 2007 11:25 AM
To: Scott Kipp;
Subject: RE: [HSSG] Soliciting Support for my Presentation

With regard to the photo, I am glad you saw the humor although I fear you may have misinterpreted some fine points it held. Perhaps you are seeing things from a different side of the curtain?
I will correct a few points you made.
First off; "The 100GE ONLY camp wants to set the standard the way it's always been done with a 10X leap and that's the only way they will accept it." I hate to say it, but I believe this statement is patently false. The issue is not that we will "only accept a 10X leap", but rather that a 10X leap has been demonstrated to meet all PAR and 5 criteria elements by a super-majority of the HSSG and 40G has not.
Secondly; I have not said that 40G PMDs have not been shown to be technically feasible. My issue is with "Economic Feasibility".  To put it simply, if we can get 4X the performance for nX the cost, that needs to be demonstrated and then as a group, the HSSG needs to determine whether that justifies the investment in a standard. Thus, I have not denied an obvious truth, I have defined one. Perhaps you are denying one?
Thirdly; The concern about whether or not 40G might/might not get a PAR if they took an alternative strategy of standing on their own two feet rather than using the 100G PAR as a hostage is worth discussion. How should we proceed in the standards process? Should we tie alternative technologies together and demand that they be approved together, or let them stand on their own? You might think one way, I might think another. We can discuss it and as a group make that decision. I am simply trying to get the elephant in the room identified for what it is, and not pretend that it is something else.
When we talk about "Distinct Identity" in a Project Authorization Request, we are telling the IEEE that the solution we are proposing will stand by itself and that it will not interfere with alternative standards projects. There are many fine lines that can be used to draw distinction, and these lines should be discussed and agreed upon. But to deny that a line exists serves nobody. There is absolutely zero doubt that 100G, being 10X faster than 10G, can be defined as distinct. I think there is a line between 40G and 100G that is much harder to distinguish and thus believe we should explore it rather than ignore it. The line between 40G and 4x10 LAG is even finer. You may disagree, this is what the process is all about. So, to specifically respond to your question, regarding frazier_03_0507, yes, there are objections to the presentation. Simply making a declaration of distinct identity does not make it so. In addition, making declarations with regard to the performance/cost ratio being better than 4x10 LAG without quantification, does not demonstrate economic feasibility. My presentation dove_02_0507.pdf shows that 4x10LAG can be a lower cost solution than 40G and that the performance difference has not been quantified. Lets quantify it and then determine if it merits the cost/investment in a standard by presentation, motions, and by following a process that is defendable.
In addition, if you look at my presentation dove_01_0507.pdf you will see that the HSSG has received many-many presentations on the economic and technical feasibility of 100G PMDs and more importantly, motions declaring that the HSSG has determined economic and technical feasibility for 100G PMDs, in addition to all of the other motions that are necessary to establish group concensus on the PAR. Look at the record for 40G and you will see that the group has not received sufficient presentations, nor made/affirmed motions in support of 40G PAR components. You show me the motions, and I will relinquish this point. Otherwise, I hope your eyes are opened to my concern that 40G has not been proven to meet the 5 criteria.
You are right that this is the HSSG, and not 100G Ethernet study group. If you inspect the record, you will see multiple motions/straw polls that show the group had centered on 100G and established that a 40G PAR should not delay 100G. Unfortunately, this clear direction by the HSSG has been violated by the events of our last meeting.
Confusion in the market - Different PMDs? Perhaps this is a way for 40G to obtain unique identity. I think that different PMDs would have helped a lot, but adding OTN signaling rates to a backplane or clustering PMD is incongruous at best, and perhaps a bit deceitful at worst. What do you think about that? Assuming the PMDs are focused on server connect, how do we ensure that a 10Km PMD or a 40Km PMD do not spring up later and create the much broader market confusion we are concerned about? Has the acceptable level of market confusion of having two higher speed Ethernet standards developed simultaneously been explored and agreed upon?
Regarding super-majority and super-minorities and the IEEE rules... these are the rules. How we use them, and what our motivations are for using them in a particular way are of interest. I spoke at the  meeting in support of the rules. I understand them and appreciate the value proposition they offer. I was addressing the motivation for using them in the way I perceived them being used, to essentially authorize an unproven project by holding a proven project hostage.
I understand your apparent concern about 40G not being able to stand on its own two feet. I understand why the 40G camp would take this tactic, but I do not condone it.
100G Ethernet clearly meets the 5 criteria, and 40G *might* meet the 5 criteria, but this has not been shown.
Rather than continue to belabor the issue, I think we should figure out a path for resolution that does not include holding one PAR or the other hostage. There may be such a resolution and I would certainly prefer to see it developed, rather than continue in the stalemate we have today. The stalemate serves nobody.
I have a few ideas for how to move forward rolling around in my head. I will be happy to share them when they have been fully formulated and will do this as soon as possible.