Can Intel Make Transceiver Peace?
Big deal? Could be, particularly if Intel can use the announcement to help speed resolution of an ongoing standards dispute.
Let's take it from the top: Intel says its new TXN17201/9 transceiver is designed to snap onto adapter cards that fit into switches, servers, storage devices, and other data-center gear, supporting interdevice links of 10-Gbit/s at distances up to 65 meters. A singlemode fiber version is slated for release "very soon," the vendor says.
The news is important on several fronts. First, it exemplifies a growing trend to faster data rates in enterprise networks. According to Intel, links among servers, storage devices, and desktop computers are getting faster, to keep up with growing bandwidth demands. Indeed, market research indicates that enterprise applications will account for the lion's share of growth in the optical transceiver market through 2006:
Intel's announcement also hints at a possible timetable for widespread adoption of 10-Gbit/s Ethernet services. According to the company, enterprises want faster desktop connections, and that's leading them to look at 10-Gbit/s Ethernet to link data-center devices. It's just a matter of time before carrier links coming into the enterprise to support those devices catch up.
"We see Intel shipping 1-Gbit/s desktop adapters in volume this year," says Bob Zona, senior product marketing manager at Intel. The history of higher-speed adoption among Intel OEMs, he says, signals that 10-Gbit/s Ethernet services should take hold in late 2003 or early 2004.
Zona points out that while optical transceivers herald faster links between devices, there's also a need for efforts to make buses and backplanes within servers, storage gear, and desktops faster and more efficient (see New Forum Develops Backplanes). Those efforts, too, will be key to the widespread deployment of 10-Gbit/s Ethernet, he says.
But the real significance of Intel's announcement may lie in its potential to help force resolution of an ongoing standards struggle. That's because the new transceiver is among the first to support the specs set by the XPAK Multi-Source Agreement Group. It's also beaten to market X2, a competing standard.
Both XPAK and X2 are newer, leaner form factors based on Xenpak, a popular design for 10-Gbit/s optical transceiver chips (see Is Xenpak Past It?, XPAK Group Unveils Specs, and X2 MSA Launched).
The goal of XPAK and X2 was the same: namely, to streamline the chips needed to support 10-Gbit/s transceivers without adding the cumbersome laser-cooling options associated with Xenpak. (Lots more information on such multisource agreements is included in Light Reading's latest report: 10-Gig Ethernet Transponders).
But the implementation details have proven divisive. Intel, Infineon Technologies AG (NYSE/Frankfurt: IFX), and Picolight Inc. spearheaded XPAK, along with BlazePhotonics Ltd., Kodeos Communications Inc., Molex Inc. (Nasdaq: MOLX/MOLXA), Network Elements Inc., and others.
X2, which was formed after the establishment of the XPAK group, is supported by Agere Systems (NYSE: AGR), Agilent Technologies Inc. (NYSE: A), JDS Uniphase Corp. (Nasdaq: JDSU; Toronto: JDU), NEC Corp. (Nasdaq: NIPNY), Optillion AB, and several others.
XPAK supporters don't like the emergence of a competing spec. "It's hard to understand," Zona says. "We were pretty surprised to see something separate emerge from essentially the same technical proposal."
But emerge it did, much to the chagrin of customers -- at least according to Zona. "We are getting tremendous pressure from the customer community for one answer, and we have made some progress," he says. But it's difficult, "for a lot of reasons."
The differences between the two approaches are minor, at least from Intel's perspective. Sources at X2 did not respond to requests for comment at press time. But according to Intel's Zona, the chief difference between the two specs lies in how they fit onto the customer's adapter card, achieving the all-important "hot swappability" customers demand. (Speakers from Agere and Agilent talked about MSAs in a recent Light Reading Webinar. Check it out in our archives).
While both have tiny rails that enable them to slide onto customer modules, X2, Zona concedes, may have a bit more flexibility in terms of the height of the rail that attaches to the adapter. "Although why that's an advantage, I'm not really clear."
Another source, Warner Andrews, VP of marketing at Picolight, says X2 doesn't support the Intel-compatible PCI bus. "We think that robs them of a high-volume market," he says.
Andrews thinks nearly all the X2 players would be willing to join XPAK, but one or two may have a product agenda that would make it more advantageous to have their own separate implementation spec. Just what that is, he isn't sure.
If Intel can use its first-strike advantage to convince X2 supporters to return to the XPAK camp, that will no doubt add to the importance of this announcement. Until then, it remains an interesting addition to the growing chorus of support for greater capacity and higher bandwidth support in enterprise gear.
— Mary Jander, Senior Editor, Light Reading