10-Gig Ethernet Transponders

One of the key reasons that Ethernet has had such phenomenal success in enterprise networks is that vendors have gotten behind standards, and as a result, they’ve managed to drive up volumes while driving down costs. Thus, everybody -- users, system vendors, and component makers -- have wound up getting a great deal. Efforts to repeat this success at the next notch up in Ethernet speed (10 Gbit/s) are now under way, with the prospects of this technology having a big impact on telecom as well as enterprise networks. This is evident from high level of activity among developers of 10-Gbit/s Ethernet transponders, or "modules" as they’re often called. These developers have been thrashing out a collection of multi-source agreements covering different applications, and they’re beginning to ship products meeting some of these specs. The time has come, in other words, to come to grips with this latest development in transmission technology. This report aims to provide a starting point by setting the scene and identifying key transponder trends, reviewing requirements for 10G Ethernet transponders, and examining the major multi-source agreements. It also presents a table of vendors and products in the market. Here's a hyperlinked summary of the report: A preview of this report, in the form of an online presentation by the author, Simon Stanley, Principal Consultant at Earlswood Marketing Ltd., can be downloaded free of charge from Light Reading’s Webinar archive.

Introduction by Peter Heywood, Founding Editor, Light Reading

1 of 10
Next Page
Page 1 / 2   >   >>
[email protected] 12/5/2012 | 12:51:04 AM
re: 10-Gig Ethernet Transponders The author stated that the 850 nm serial PHY supports 300 meter links over MMF, but this is only the case if you have special 2000 MHz*km, 50 micron core MMF. This fiber is described in TIA-492AAAC, but I think this is very new and virtually none is installed anywhere. Normal 62.5 micron MMF can support links of only about 30 meters with the 850 nm serial PHY.
To support the overwhelming installed base of 62.5 micron MMF, the Task Force came up with the 1310 nm 10GBASE-LX4 PHY, which is 4 wavelengths (1275, 1300, 1325 & 1350 nm) each modulated at 3.125 GBits/sec and mux'd/demux'd onto/off of a single fiber using CWDM or WWDM techniques. Link lengths of 330 meters have been demonstrated by the Molex, Pine & Blaze NP consortium (maybe as part of the old XENPAK MSA group). This supports 85% of LAN backbone lengths, is compatible with the 85% of fiber in the metro being MMF and is consistent with 85% of 1 GigE optical ports shipping with SX (for MMF) rather than LX (for SMF), which is why the Task Force selected the LX4 PHY.

The full details of fiber types supported for various 10GBE PHYs is at the 10 Gigabit Ethernet Alliance web site (www.10gea.org) in a whitepaper entitled "Optical Fiber and 10 Gigabit Ethernet".
Yellow Jacket 12/4/2012 | 11:17:26 PM
re: 10-Gig Ethernet Transponders To all and the author of "Misleading/erroneous facts by Author"

The "special 2000 MHz*km, 50 micron core MMF. ..described in TIA-492AAAC" that supports low cost 10GBASE-SR to 300 meters is installed widely in networks, with over 100 million meters installed in premises networks since 1999." That is a fact. And customers installing this fiber, also known as OM-3, are just the same customers thinking about 10G Ethernet upgrades. These customers will enjoy low cost 1000BASE-SX for 1 Gig, and low cost 10GBASE-SR at 10 Gig. And customers installing 10G Ethernet today that have existing conventional mulimode fiber can save thousands by using the OM-3/SR comination. They will spend a few hunded hundred dollars per link to pull in OM-3 cable, and save thousards per link by using 10GBASE-SR instead of LX4 or LR.

This is not to say there is no place for LX4 or other solutions that run over conventional MMF. It may make sense to spend the extra premium for these modules and use the expensive mode conditioning patch cords required by the 802.3ae standard, for networks in which cable replacment is unusually expensive.
DiamondDave 12/4/2012 | 9:53:26 PM
re: 10-Gig Ethernet Transponders Great synoptic report, but...

Who's buying volume of any 10GbE transponder? The tech side is covered well, but what about the market side?

With Infineon and Agere out of the opto biz, Nortel and JDSU teetering on the brink, start-ups like OptronX (et al.) needing cash via sales or funding ASAP, carriers continuing to reduce capex, 2002 a near write-off and 2003 looking marginal (at best)...it would seem that the tech side is secondary.

The big questions: What tech will prevail? Who will buy it? And specutively, who will be around to produce it?

Alas, it was great to have everything collectively laid out - but is there a follow-up planned?

On a semi-related note - in the SONET/SDH world, do you see demand or interest for discreet 10G transmitters and receivers?
Simon_Stanley 12/4/2012 | 9:53:15 PM
re: 10-Gig Ethernet Transponders For companies selling hot pluggable transponders such as XENPAK there is likely to be delay in shipments not only because of the market slow down but also because equipment manufactures can ship equipment without all the transponders slots populated. This also makes it much more difficult to forecast accurately.

The good news is that XENPAK transponders and 1310nm lasers with 40km reach will significantly reduce the cost of 10G Ethernet solutions and may accelerate adoption.

As to who will buy 10GE transponders then I think the biggest purchasers will be those in the enterprise and metro ethernet markets.

Who will be around to supply transponders in 1-2 years? I think this is difficult to answer and I am sure we will see several more companies, both big and small, pull out. Agere was a sponsor of the Webinar 6 days before their announcement to pull out of the opto market. Infineon at the moment is fully committed to this market.

We are discussing a follow-up where we would look more at the market. If you have any ideas then please send them in.

Finally, on the SONET/SDH side I personally believe the market for discrete transmitters and receivers will be very limited as the cost and features makes them uncompetitive with transponders.
optodunce 12/4/2012 | 9:53:11 PM
re: 10-Gig Ethernet Transponders Simply nuts!!! So I get it if 80 % of all desktops use Ethernet as a platform...of course the natural migration is lets extend it to the CO and then between CO's then Between Cities and then around the world!

Simply Bizarre...Has anybody posed the question should we migrate a technology that is based on a shared enviornment!!!? As it is even as a private LAN, Ethernet is a un-secure network reliant on encryption and software heavy security that can easily be accessed...but to extend it to the outside plant is ludicrous!

We have a very effective design in place utilizing copper, POTS, instead of taking the POTS design and overlaying an optical network that emulates it(not PON), we extend a foolish design rich in protocols and layer it on top of risky shared network enviorments, and then try to squeeze more speed on this risky network!

Ethernet will prove to be an in effective design for provider applications...and to follow down the ethernet road is just like following down the 1022x 1022 MEMS road Xros promoted two years ago!!! Where is that technology...as the bottom of the St Lawrence!

fiber_r_us 12/4/2012 | 9:53:08 PM
re: 10-Gig Ethernet Transponders optodunce, you are living up to your alias here:

- Ethernet stopped being a shared media when Ethernet switching was invented over 10 years ago.

- Ethernet is used by 99+% of all desktops and servers.

- Using Ethernet between packet switching devices does not require encryption and is no more and no less secure than any other fiber-based link technology.

- Copper and POTS is a "very effective" design for what? Voice?? It clearly sucks for data!
beowulf888 12/4/2012 | 9:52:00 PM
re: 10-Gig Ethernet Transponders Well, at the risk of igniting another bell-head/net-head flame war (we're like two feuding tribes, aren't we?), I have to ask you, Optodunce, have you been asleep for the last decade? Data networks *have* evolved, you know. I remember an "expert" from PacBell telling me back in 1991 that a 56K channel was enough for any "properly designed data application". That wasn't you by any chance?

Anyway, from a standpoint of CSMA/CD, Ethernet is a shared medium. But now that most Ethernet is deployed in a switched environments, it's no more "risky" than any circuit-switched network.

You might feel threatened by "designs rich in protocols", but last time I looked the ITU was contributing their fair share to the protocol soup. Compartmentalizing functionality in descrete layers (and discrete components within each layer) is a tried-and-true methodology. Remember why they developed SS7 -- to improve security by separating the signaling from carrier.

BTW: comparing MEMS, a technology, to Ethernet, a set of protocols, doesn't make sense to me. An analogy would be TCP/IP over carrier pidgeon. If IPoCP doesn't work very well, it's not the protocol's fault. It's the fact the medium (the "technology") is slow and kind of dirty. ;-) If the 802.3ae initiative fails, it won't be because Ethernet as a protocol is somehow flawed.

Personally, I doubt if 10GigE will be as big as its proponents predict. CapEx budgets are tight, and the Telcos will be writing off their current SONET infrastructure well into this coming decade. Certainly 10GigE technology is cheaper than SONET if one were to consider the purchase price only. But if one were to consider the OpEx costs of bringing in a new technology, 10GigE looks somewhat less attractive. Still this doesn't mean the technology is in anyway inferior -- it just might not make economic sense at the moment.

Well, it just ain't your father's Bell System anymore...


Michigan 12/4/2012 | 9:51:51 PM
re: 10-Gig Ethernet Transponders
Not only was Intel not a founding member of the 300-pin MSA (as this article claims), Intel is not even a member of the 300-pin MSA.

pavelz 12/4/2012 | 9:51:39 PM
re: 10-Gig Ethernet Transponders Not to read too much from a name, but IMHO the term transponder is abused here, and it's somewhat mixed up with "MSA".

Not that I would know of a better name; do you?
That is, a name for

"Pluggable optical (receive/transmit) module with some electronics." (POMSE! my own acronym, not understood by anyone!)


If you want an argument, here is why I think transponder is not right:
these modules are not really transponders because while it's true that the signal goes in at one rate/coding and out at another one, for each T and R, from the point of the system, this is just an implementation detail (IEEE says that XAUI is optional). It does NOT transpond Ethernet optical to Ethernet electrical, say; it does much less than that (from the system point of view).

Another point against usig 'transponder' for 'these modules': XFP can not be properly called a transponder; but it in fact it compares to XENPAK for most considerations at the system level purposes ("are the optics a pluggable module?", etc).
In this article the point is noticeable - the term MSA is used in the end. That is also a poor term; fro a point of view of a system designer, who cares if it's multisourced agreement, or ad-hoc, or whatever.

lightflash 12/4/2012 | 9:51:04 PM
re: 10-Gig Ethernet Transponders LightReading has omitted Molex 10Gig modules.
Molex is currently shipping LX4 under MuxLink brand name. Also Molex will have XENPAK and XPAK LR transponders in Q4.
Page 1 / 2   >   >>
Sign In