| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Thanks Brad, good summary From: Brad Booth [mailto:Brad_Booth@xxxxxxxx] Jonathan, Paul summarized the issue very well: “The trouble with AOCs is that if port-lock-out policies are in force, both ends of the channel must plug into the same brand of switch or server. That is an unattractive constraint customers face with surprise at first followed by bitterness. They fault IEEE for not doing its job to ensure interoperability.” The same thing is occurring with SFP+ modules. The industry is deploying their 2nd and 3rd generation of equipment with SFP+ and are trying to re-use SFP+ components that support port-lock-out. To the customer, they don’t know why it doesn’t work. They have an SFP+ socket and an SFP+ component. The component used to work in the previous equipment but doesn’t work in the new equipment. As Paul says, they fault us for not doing our job to ensure interoperability. As for what we can do, the study group/task force cannot ensure interoperability. Companies can produce proprietary implementations, even if they are based on industry standards or MSAs. In my humble opinion, if there was a way to address this, then I’m sure the industry would be quite happy. Thanks, From: Jonathan King [mailto:jonathan.king@xxxxxxxxxxx] Hi Brad, Could you summarize the interop issues, and whether you think they could be addressed by IEEE or MSA? Thanks jonathan From: Brad Booth [mailto:Brad_Booth@xxxxxxxx] Jeff, Brad, I was sincerely asking what needs to be further standardized to make AOC work for end customers. The customer has nothing to choose. There is no mix and match scenario. It is “pluggable form factor to cable to pluggable form factor” that is all permanently connected. There is nothing to configure in a physical sense, only software. In relation to software, one barrier I see is something we do not cover in 802.3 and that is the required digital diagnostics such as transmitted and received optical power. AOC tends to eliminate this function since mimicking copper cabling where as transceivers are best when offered with such functions. Hence, the customer is confused as to why link health monitoring based on optical power is not consistently provided. I could attend the maintenance meeting, but perhaps you could give us a hint as to one of the big ticket problems. SFP+ is confusing with some modules being limiting and others being linear and with system companies possibly implementing support only for limiting modules. The linear SFP+ modules fail in these limiting slots. None of this though as to do with IEEE because electrical interfaces are optional. Should they become mandatory specifications in the IEEE? Jeff From: Brad Booth [mailto:Brad_Booth@xxxxxxxx] Jeff, Brad, The CAUI or XLAUI electrical interface is a 802.3 standard that these AOC’s connect. What more would we need to “standardize?” AOC can go 100 meters or more physically. It is market acceptance of such assemblies for longer runs that is the problem owing to installation difficulties. We still need standards for transceivers (pluggable optics) but I think it is difficult to preclude the adoption of AOC that are based on whatever novel technology meets the need as in cost as things move along. There are compliance points that a transceiver based link has to meet that an AOC can ignore. Perhaps what should be looked at is easing the compliance points for Ethernet optics while keeping the link budget stable. Jeff From: Brad Booth [mailto:Brad_Booth@xxxxxxxx] I liked Chris’ point about 30 m vs. 10 m. While the study group may decide on two objectives, it doesn’t mean it has to be met by two PMDs. This was the case in P802.3ae where there was a 2 km SMF objective that was satisfied by the 10 km SMF PMD. The P802.3ae task force felt that there was going to be very little cost difference; therefore, it was better to put the volume into one PMD type rather than segment the market for with two PMDs. This could be a similar situation for MMF PMDs. It may be easier for the eventual task force to have the ability to make those trade-offs between providing one or two PMDs to satisfy the different market requirements. The one requirement I perceive as point-to-point leaf-spine or EoR/ToR deployment HFT, HPC and Web 2.0 applications, and the other requirement is structured cabling data center environment (either Enterprise or Cloud). As for AOCs, I believe they are a fantastic example of missing the broad market potential. Why do customers use AOCs? Ease-of-use? Not likely. Cost? Definitely. If the study group decides it wishes to concede the shorter-reach, cost-optimized market to a non-standard implementation, then in my humble opinion that brings into question the viability of broad market potential for the other PMDs. Why would the study group not pursue what is potentially the largest volume market with an IEEE 802.3 standard? Just my 2 cents, Brad From: Mark Nowell [mailto:mnowell@xxxxxxxxx] I think Chris has some good points here. I can see the argument for the reaches being bi-modal with 100m being one maximum and a second one at a shorter reach (I hear 30m, 50m, 70m). This is resulting in many assuming we need two PMDs. I don't see it that way. Chris' points below around the challenges of doing two PMDs are all worth considering. I do want to raise what I had assumed would be the market approach here: - standardize around 100m - assume AOC addresses any cost-optimized shorter reach. From a standards perspective, we define the 100m PMD, we also define the 4x25G electrical interface. AOC's are interoperable by the standardization of the electrical interface. The advantages of AOC's from an application perspective, is that we are all aware of the cost sensitivity that very short reach has. An AOC allows a supplier to optimize their technology in any way they choose. In fact, it doesn't limit technologies to just MMF ribbon. It allows the user to focus on the 3 things that really matter - cost, power, size.... Question for those thinking about two PMDs. Is the assumption that these two PMDs are in anyway interoperable or are they distinct? And which two? I've heard 100m OM3/150m OM4 and I've heard 100m/~50m OMx. Mark On Feb 24, 2012, at 12:14 PM, Chris Cole wrote: Steve Last year, I was on the same page as you, in viewing two MMF objectives as appropriate for supporting two distinct applications. One application is lowest possible cost for shorter reach applications, for example 50m as you suggested, or 70m as I suggested. A second application provides broad data center coverage, for example minimum of 100m. It is also my perception that there is broad agreement in the Study Group about the existence and importance of these two applications. However, I have now come around to the view point that we should move forward with only one MMF objective, specifically minimum 100m on MMF. Technical solutions as to how to achieve this objective such as pre-emphasis, DFE, other EDC techniques, FEC, OM4, should be left up to the Task Force to decide. The reason I no longer support the shorter reach objective is that I do not see the possibility for consensus as to what the technical solution will look like, nor do I see us as having enough technical information to make informed trade-offs. So it is in the best interest of the low cost application not to prematurely freeze design choices by adopting a standard. We do not have enough information about VCSEL yields to optimize for cost. Any low cost reach objective we pick will either be shorter than necessary, or longer than optimum. Design choices like un-retimed interface which will be more feasible a few years from now, are likely to be off the table now.
When we have learned more, we can adopt a second 100G MMF standard in a future project, if it still makes sense. Chris From: Swanson, Steven E [mailto:SwansonSE@xxxxxxxxxxx] Pete, Not really; the resulting PMD was 100m over OM3 and we got to 150m on OM4 by reducing the connector loss from 1.5 to 1.0 dB. We cannot do that here unless one would define only a 150m PMD and as Petar notes, this likely would not fly because the 150m PMD by definition will be higher cost than a shorter reach PMD and is not what is needed by some. One size does not fit all. Steve Steven E. Swanson t 828-901-5328 From: Anslow, Peter [mailto:panslow@xxxxxxxxx] Steve, Well, 802.3ba is a counter example. The objective was 100m and the resulting PMD was 150m over OM4. Regards, Pete Anslow | Senior Standards Advisor From: Swanson, Steven E [mailto:SwansonSE@xxxxxxxxxxx] Pete, If we set the objective at 100m, we will not get 150m. In fact the argument in the TF will be “we don’t need it to meet the objective.” At least that is my experience in several of these developments. Best regards, Steve Steven E. Swanson t 828-901-5328 From: Anslow, Peter [mailto:panslow@xxxxxxxxx] Paul, While I am not saying that 150m over OM4 may not be the outcome of the Task Force, it is my view that it is not the function of the Study Group to start to make technical choices for the Task Force. I think the Study Group should set an objective at which the PMD has broad market potential etc. and try to make as few technical choices as possible. 150m over OM4 is quite a challenging objective and setting it would rule out some of the choices that the Task Force might want to make. I am not of the view that if it turns out to be 100m over OM4 that this is not worth doing, so I am more in favour of setting the objective to be 100m over MMF and leave the technical choices as to whether this means over OM3 or OM4 etc. to the TF. Regards, Pete Anslow | Senior Standards Advisor From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxxx] Pete, The line of thinking expressed in your second paragraph suggests that the MM reach objective would need to be revisited, presumably to be shortened for lower cost, should the SM PMD cost be projected to be low enough to cause it to be attractive for channel lengths within that MM reach. If my interpretation is correct, and we do not have confidence in SM optics achieving such an aggressive cost decline, then the reach objective for MM should be at least as long as that established for 100GBASE-SR10, namely 150 m on OM4. Regards, Paul From: Anslow, Peter [mailto:panslow@xxxxxxxxx] Chris, I don’t think that this really helps. What you really need to know in order to determine if an SMF PMD will affect what the reach should be for the MMF objective is the relative cost of the new MMF and SMF modules. However, just knowing what the SMF reach objective is won’t tell you that. Many of the decisions to be made by the Task Force will affect this relative cost and there was some consensus on the last SMF call that the only likely constraint on an SMF objective below say 2 km is the cost of parallel fibre. It is not clear at this point whether a PMD capable of 1km (say PAM-8) is more expensive or cheaper than one that is only capable of 500m (say parallel fibre). I think that the Study Group should decide on a MMF objective that stands on its own. If it turns out that there is an SMF objective and the Task Force choices make the relative cost of that PMD low enough to affect the MMF reach, then the Task Force can seek to change the MMF objective. This is (in my opinion) the earliest point that the group can rely on having a reasonable idea of the relative cost of the two solutions. Regards, Pete Anslow | Senior Standards Advisor From: Chris Cole [mailto:chris.cole@xxxxxxxxxxx] Steve I tend to agree with you, but Paul does have a point that to some knowing what the SMF objective is important in deciding what the MMF objective should be. My suggestion is that we keep the MMF poll exactly as Jonathan crafted it. Anyone that feels the SMF objective has to be decided first has the option to state so and then state their view of what that is likely to be. This will allow us to not delay making progress towards reaching consensus on a MMF objective So let’s call this question A. A) To make a decision on an MMF reach objective, I am assuming the SMF reach objective will likely be: a. No SMF objective b. At least 500m c. At least 1000m d. At least 2000m Those like you and me who do not see strong linkage simply do not answer question A. Jonathan can then record how many responses he receives to question A and the choice. Both results will give us a measure of the thinking of Ad Hoc participants. Chris From: Trowbridge, Stephen J (Steve) [mailto:steve.trowbridge@xxxxxxxxxxxxxxxxxx] Hi Paul, I don’t think it is nearly so clear that you should decide SM first. If there existed a SM solution that was cost-competitive with a MM solution at some reach, it would be a game changer, and those developing MM solutions would surely like to know if the game will change before getting too far down the path. But most seem to believe that the game will not change, and even if it did, it is hard to prove because it is difficult to compare relative costs of dissimilar technologies. Most still seem to believe (in spite of the interesting technology from Opnext) that there will be a significant step function from MM to SM that will keep people from wanting to use it in data centers. Furthermore, if you need a different cable type, for example, for a 70m link than you need for a 100m link, that creates its own kind of problem. So if the game does not change, then SR4 needs to try to address most, if not all, of the reach currently addressed by SR10. If it turns out not to be technically or economically feasible to do that (e.g., if you could only get 60 or 70m out of the beast within reasonable cost, size and power), then if SM is to replace MM above that reach, it needs to get down to a cost to compete with an SR10 with a reverse gearbox. Even if SM does this, it isn’t clear they will get all of that market because of a likely reluctance to mix cable types in the data center – maybe they are happier to use SR10 with a reverse gearbox to reuse their existing cabling. Regards, Steve From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxxx] Jonathan, While I understand the desire to find a launching point for the discussion, this poll is approaching the problem in the wrong order because we need to know what the single-mode objective is first. A purpose of the Study Group is to set objectives that will allow us to establish cost-optimized 100GE. One cannot logically pick cost-optimized MM objectives without first having framework around the SM optics that will be used to address channels with lengths that exceed the MM reach. At this point, we have not even established if we will have a SM objective. In other words, we don’t know if the existing LR4 will remain the only one, or if there will be another one added. Only when the SM situation is established can we know the minimum capability that a new MM optic must fulfill to optimize cost. I suggest that we first conduct such a poll for SM and use it to start the objective discussion in the SM ad-hoc. If that produces solid results, then undertake the same endeavor for MM. Regards, Paul From: Jonathan King [mailto:jonathan.king@xxxxxxxxxxx] Dear all, On the Feb 14th MMF ad hoc call , it seemed like we were beginning to converge on a possible objective for MMF . In the next meeting (Tuesday 28th Feb), I’d like to see if we can finalize a strawman MMF objective. To that end I’ll prepare a presentation which we can review during the call which will include a strawman objective for review on the call, together with an overview of how it addresses the 5 criteria – to help get the best starting point for that discussion I’d like to get your responses to the questions below questions: The strawman objective will follow the wording in Anslow_01_0111 Define a 4-lane 100 Gb/s PHY for operation over OMX MMF with lengths up to at least Y m 1) A reasonable MMF reach objective (i.e. the value of Y) would be a. 100m b. Significantly less than 100m (what reach?) c. Significantly more than 100m (what reach ?) d. decided in the task force 2) The MMF type should be a. decided in the task force b. OM3 c. OM4 d. at least as good as OM4 Please send your responses to me directly at: jonathan.king@xxxxxxxxxxx I will collate and report the results but will not reveal any individual’s responses. If you feel uncomfortable expressing an opinion, say so and I’ll note that. To repeat, this is not a formal poll or vote, just intended to give us the best starting point for discussion on Tuesday. Please send your responses as soon as possible, and at least by close of business on Monday 27th Feb, 2012 Many thanks ! Jonathan King MMF ad hoc chair, Next Gen 100G Optics From: Anslow, Peter [mailto:panslow@xxxxxxxxx] Hi, Following on from the meetings on 14 February, Jonathan and I are planning to hold an SMF Ad Hoc meeting immediately followed by an MMF Ad Hoc meeting (1 hour each) starting at 8:00 am Pacific on Tuesday 28 February. If you would like to present a contribution at the SMF ad hoc, please send it to me and for the MMF ad hoc, send it to Jonathan. Peter Anslow from Ciena has invited you to join a meeting on the Web, using WebEx. Please join the meeting 5-10 minutes early so we may begin on time. +44-203-4333547 (United Kingdom) 4438636577 (United States) Conference Code: 207 012 5535 Australia, Melbourne : 0383380011 Australia, Sydney : 0282386454 Austria, Vienna : 01253021727 Belgium, Brussels : 028948259 Bulgaria, Sofia : 024917751 Canada, All Cities : 2064450056 China, All Cities Domestic : 8008706896 China, All Cities Domestic : 4006920013 Czech Republic, Prague : 228882153 Denmark, Copenhagen : 32727639 Estonia, Tallinn : 6682564 Finland, Helsinki : 0923193023 France, Paris : 0170375518 Germany, Berlin : 03030013082 Germany, Frankfurt : 06924437355 Hong Kong, Hong Kong : 85230730462 Hungary, Budapest : 017789269 India, Bangalore : 08039418300 India, Chennai - Primary : 04430062138 India, Mumbai : 02239455533 India, New Delhi : 01139418310 Ireland, Dublin : 015269460 Israel, Tel Aviv : 37630760 Italy, Milan : 0200661900 Japan, Tokyo : 0345808383 Korea (South), All Cities : 0264903634 Latvia, Riga : 66013622 Lithuania, Vilnius : 52055461 Luxembourg, Luxembourg : 20881245 Malaysia, Kuala Lumpur : 0348190063 Netherlands, Amsterdam : 0207946527 New Zealand, Auckland : 099291734 Norway, Oslo : 21033950 Poland, Warsaw : 223070121 Romania, Bucharest : 318144966 Russian Federation, Moscow : 4992701688 Singapore, All Cities : 6568829970 Slovak Republic, Bratislava : 0233418490 Slovenia, Ljubljana : 016003971 Spain, Barcelona : 935452633 Spain, Madrid : 911146624 Sweden, Stockholm : 0850512711 Switzerland, Bellinzona : 0912611463 United Kingdom, All Cities : 08443386571 United Kingdom, All Cities : 02034333547 United States, All Cities : 4438636577 Vietnam, Ho Chi Minh : 84838012419 Regards, Pete Anslow | Senior Standards Advisor |