Dale’s Web Log

After months of internal squabbling, members of the Institute of Electrical and Electronics Engineers Inc. (IEEE) Higher Speed Study Group (HSSG) have finally agreed on a path to the next standard for high-speed Ethernet.

At a meeting last week, HSSG members approved the project authorization request (PAR) and the IEEE’s “five criteria” required for the study group to continue working on a new Ethernet standard that will include both 40Gbps and 100Gbps rates.

The agreement comes after six months of internal debates over whether the next standard should include a 40Gbps line rate, and a full year after the study group was founded.

The new standard will provide physical layer (PHY) specifications to support 40Gbps operation over at least 100 meters of multimode fiber, at least 10 meters of copper, and at least 1 meter over a backplane.

On the 100Gbps side, the standard will address distances of at least 10 km and 40 km on singlemode fiber; at least 100 km on multimode fiber; and at least 10 meters over copper.

The HSSG was formed to define the next Ethernet standard for network aggregation, and it seemed well on its way to advancing a 100Gbps proposal late last year.

But in January some members of the group, led by server and storage vendors such as Sun Microsystems Inc., Hewlett-Packard Co., and Brocade Communications Systems Inc., began pushing for a standard that included a 40Gbps rate for data-center applications.

A number of switching and routing vendors, including Cisco Systems Inc., rejected that idea, citing possible delays in the commercialization of 100-Gigabit Ethernet aggregation products, as well as the higher costs involved in developing and manufacturing products that comply with a dual 40- and 100Gbps standard.

In the end, the group was able to secure a large enough majority to adopt the proposal for a dual standard. In the five-criteria document, the HSSG noted that “bandwidth requirements for computing and core networking applications are growing at different rates, which necessitates the definition of two distinct data rates for the next generation of Ethernet networks.”

Specifically, a majority in the group decided that 40Gbps provided “the best balance of performance and cost” for servers and computing applications, while 100Gbps was the better speed for aggregation and core networks.

On the HSSG email reflector on Friday, study group chair and Force10 Networks Inc. components scientist John D’Ambrosia wrote that the IEEE 802.3 Executive Committee had approved the pre-submission of the group’s PAR to the New Standards Committee (Nescom) for consideration at the December 2007 Standards Association Standards Board (SASB) meeting, and that it would remain on the agenda subject to Executive Committee approval at the November meeting.

D’Ambrosia also noted that the Executive Committee had approved an extension of the HSSG, allowing it to continue working on defining the standard.

Brad Booth, president of The Ethernet Alliance , says that, despite the internal debates, he doesn’t expect much delay in the standardization process. “I would expect formal ratification sometime in 2010,” he says. “The real technical meat of the work can start, and has already. We’ve seen presentations that are very technical in nature, so we may have a first draft by next summer, which would put us in line with where we hope to be”

— Ryan Lawler, Reporter, Light Reading

ANAHEIM, Calif. — OFC/NFOEC — If 100Gbps Ethernet gathers pace quickly enough, it might put a crimp in the lifespan of the 40Gbps generation.

That’s one possibility being discussed here at OFC/NFOEC, as industry executives wonder whether 40Gbps might see a shortened lifespan due to pressure from both 10- and 100Gbps alternatives.

“We see the 40Gbps deployment as more of a stepping stone,” says Saeid Aramideh, vice president of marketing for CoreOptics Inc. “Not that we have stopped our activity there, but certainly we see our future being 100Gbps-based. My personal belief is that with the coming of 100Gbps transmission in the WAN, the 40Gbps life cycle could be short-lived.”

Metro and long-haul 100Gbps deployments are years off — most sources are saying 2012; AT&T Inc. has suggested 2010 — while 40Gbps deployments are underway now. AT&T has lit its OC768 backbone, and here at OFC/NFOEC, Verizon Communications Inc. officials said they also plan to build a 40Gbps core.

But here’s the catch. It’s generally accepted that for 40Gbps sales to take off, enabling 40Gbps to usurp 10Gbps, the cost should be no more than 2 to 2.5 times as much as 10Gbps. So far, 40Gbps prices aren’t there.

“The cost economics of 10Gbps are so strong right now, it’s limiting 40Gbps to only those cases where they have to use it,” says Roy Rubenstein, research director with the transceiver market research firm, LightCounting . A typical, short-reach, 40Gbps transceiver can carry a $20,000 to $25,000 price tag, he notes.

So, if 100Gbps optics manage to catch up by costing, say, about five times as much as 10Gbps, could that cut short the 40Gbps generation? “Depending on where 40Gbps moves, you might see an intercept point with 100Gbps, but it’s too early to tell,” says Mike Ricci, a senior vice president at JDS Uniphase Corp.

What might make that intercept point possible is the amount of attention being lavished on 100Gbps transmission. The 100Gbps name-dropping at OFC/NFOEC includes prominent vendors such as Alcatel-Lucent, Infinera Corp, and CoreOptics customer Siemens Communications Group .

“There’s a window for 40Gbps. If people get the prices right, they can have a chance,” LightCounting’s Rubenstein says.

The optics vendors pushing 40Gbps don’t appear too worried, considering 100Gbps transmission is still pretty far from reality. “If there’s a need for 100Gbps, it’ll happen, but at this point I don’t see a significant threat to the investments made in 40Gbps,” says Ed Cornejo, director of product marketing at Opnext Inc.

That doesn’t mean Opnext is ignoring the next wave, as it’s already engaging in 100Gbps laser research in its lab. On a panel at Monday’s Optical Society of America Executive Forum, Opnext CEO Harry Bosco said the tough part, when it comes to transceivers, will be finding the chips to work at that speed.

And recent M&A activity shows confidence in the upcoming 40Gbps market. Two of this week’s acquisitions — Kailight Photonics Ltd. by Optium Corp., and Kodeos Communications Inc. by Finisar Corp. — “show people are getting serious about their 40Gbps portfolios,” Rubenstein says. Kailight is shipping 40Gbps modules, while Kodeos, more of a 10Gbps vendor, uses long-haul encoding techniques that could be useful at 40Gbps, he says.

— Craig Matsumoto, West Coast Editor, Light Reading