Open warfare broke out at last week's 10-gig Ethernet standards meeting, with carriers facing off against equipment providers over optical interfaces.
The atmosphere at the IEEE 802.3ae task force interim meeting in La Jolla, Calif., was reportedly "rancorous and unhappy."
"It's a civil war," says Thomas Dineen, a meeting attendee and an independent system and ASIC design consultant in Milpitas, Calif. "And some presentations, shall we say, didn't facilitate an attitude of compromise."
At issue were the inclusion of short-reach, 850-nanometer interfaces between networking gear -- such as routers, switches, and cross-connects -- and optical networks, including those supporting DWDM (dense wave division multiplexing). Carriers want 850-nm links, because that would allow them to create inexpensive, short-reach 10-Gbit/s Ethernet interfaces over existing multimode fiber. The links could be used, for example, to connect switches in a central office.
But sources say that equipment vendors -- such as Cisco Systems Inc. (Nasdaq: CSCO), Lucent Technologies Inc. (NYSE: LU), and Nortel Networks Inc. (NYSE, TSE: NT)-- rejected solutions that supported older fiber. Instead, they backed the set of options including 1300- and 1500-nm serial links over single-mode fiber. The vendors can charge a premium for these interfaces, which will require special "boutique" manufacturing and will force carriers to adopt new, costlier fiber.
In the end, the vendors carried the day. There is at present no 850-nm option in the baseline proposal - despite its obvious benefits to carriers. And that won't change -- unless there's sufficient consensus at the next meeting to include the 850-nm options. That's left carriers steaming. "Short reach interfaces in the CO will account for two-thirds of all 10-gig applications. It was very obvious that system vendors did not want those interfaces to be cheap," said one. "They're following the money trail, trying to standardize the more expensive options. It's all about controlling the market."
Some accuse the vendors of stuffing the ballot box. "The system vendors vote together, like a cartel, to protect their interests," said the anonymous attendee.
Not true, say the vendors. They say supporters of 850-nm links wanted to include too many interface options in the standard. System vendors were merely protecting the proposal from needless complications. "Equipment vendors don't want to confuse the market with too many interfaces," says Bob Grow, a voting member who also heads up the technical committee of the 10 Gigabit Ethernet Alliance (10 GEA), a consortium that is separate from the IEEE. He says that having too many interfaces will cause vendors to present multiple solutions, diluting the value of the standard.
The optical battle promises to continue at the next IEEE 802.3ae meeting, slated for Boston in the second week of September. "There are still some objectives of the task force that aren't being fulfilled," says Dineen. "We agreed to support legacy fiber last year. Until we do so or change the objectives, there is a hole to fill."
-- by Mary Jander, senior editor, Light Reading http://www.lightreading.com
From Jay Hoge, Principal, Firestorm Ventures: I am, in general, a big fan of Light Reading. I must, however, take exception to this article. I was, until recently, a Marketing Director with JDS Uniphase, and have been attending IEEE 802.3ae meetings since last summer. I am also the clause editor for the serial PMDs.
The split at La Jolla had nothing to do with the system vendors prefering single mode installations, as your article states, but rather how many solutions for short haul multimode fiber installations would be supported. Everyone wants to support 10GB/s operation over MMF.
The two camps were the "3 PMD Camp," including most of the system vendors, and the "5 PMD Camp", including most of the VCSEL vendors and fiber manufacturers. Both camps fully support the two SMF PMD's. The argument arises over how to support MMF.
The 802.3ae Task Force adopted a set of goals over a year ago. Among these were a set of Reach Objectives. The objectives were to support 10GB/s operation:
- 1) Over SMF at 10 Km.
- 2) Over SMF at 40 Km.
- 3) Over SMF at 2 km.
- 4) Over MMF at 300 m (note that no particular MMF is specified).
- 5) Over 160MHz/Km 62.5u MMF at 100m.
One proposal, backed by the system vendors in the 3PMD Camp, would be to adopt a 4 Channel WWDM at 1310 nm PMD, proposed by Agilent. This device works over both SMF and MMF and meets both Objectives 4 & 5. For MMF operation, it requires an offset patchcord, like that of 1000BASE-LX. The system vendors like this solution because it offers a single PMD solution for any MMF requirement. They further argue that economies of scale resulting from having a single device (versus 3) covering the MMF application space will minimize the cost differential between the PMD's. Last, they suggest that very short reach (less than 100m)intra system links are outside the scope of Ethernet and can be implemented most cost effectively by using proprietary, vendor specific solutions.
The 5PMD Camp wants to include 850nm solutions in the standard; both serial and WWDM. They maintain that, since ~90% of 10GB/s links will be less than 300m, that the market will support 3 PMD types, each optimized for a more specific application. The 850nm WWDM would cover existing 62.5u installation to 150m plus up to 550m over new extended BW MMF. The serial 850nm would cover links to 30m over existing fiber and to 300m over the new fiber. The 1310nm WWDM would cover links to 300m over existing fiber. The 5PMD Camp believes that this will result in lower costs for each application space.
In reality, the System Vendors want to support as few PMD as possible because of logistical concerns and concerns that their customers will get confused by having to many overlapping options.
The VCSEL manufacturers plus the manufacturers of high BW MMF (Corning & Lucent) will be left with nothing to sell unless the addition 2 PMD's are adopted. This group blocked the inclusion of the 1310 WWDM, which they support as part of a set, rather than risk being left out in the cold. At present, the 2 groups are still deadlocked.
Given all of this, your report totally missed the boat.