10G form factors: Too much just enough?
by Stephen Hardy
If you listen to transceiver vendors, the emerging SFP+ multisource agreement (MSA) could revolutionise 10 Gbit/s interfaces. The module sector hasn’t witnessed this kind of excitement since the advent of the XFP MSA-the form factor the SFP+ is expected to replace in many enterprise applications before the XFP had much of a chance to reach the field. In fact, the SFP+ represents the seventh MSA targeted at 10 Gigabit Ethernet and other 10 Gbit/s applications. The evolution of 10 Gbit/s form factors illustrates the difficulty the module industry has had in matching technology to customer requirements as well as the influence one or two large customers can wield. It also may indicate which of the current seven have nearly run their course, which of the newer ones hold the most promise, and which form factors may never gain much market traction.
The 10 Gbit/s form factor parade began with the 300-pin transponder originally developed for SONET/SDH applications. The trouble started with the advent of 10-Gigabit Ethernet. While pundits heralded the new standard as the convergence of datacom and telecom-which still might happen-it had the opposite effect on transceivers. In addition to its multiplicity of PMDs (there are three 10 Gbit/s PMDs that cover multimode fibre, compared to one for Gigabit Ethernet, for example), 10-Gigabit Ethernet begat, in relatively short order, XENPAK, X2, XPAK, and XFP. This year has seen the emergence of XFP-e (targeted at the carrier sector) and SFP+.
Did the industry really need to develop all of these form factors? In a way, yes, says Daryl Inniss, vice president, communications components, at market research and consulting firm Ovum-RHK (www.ovum.com). “There’s actually a progression to these; they didn’t all start at the same time,” he points out, referring to a general trend toward smaller modules packing fewer electronic functions. “I don’t see it as a problem per se. It’s how the market is growing up; we’re converging onto something.”
But that convergence began with divergence, particularly between telecom and datacom applications. “There were definitely technical reasons why we have some different ones,” says Ed Cornejo, director of product marketing at Opnext (www.opnext.com). “In datacom, you had a different electrical interface that you had to deal with for 10G, XAUI [the first 10-Gigabit Ethernet standard interface]. In telecom, you have the SFI-4. And at some point, people will want to move to some sort of serial 10-gig interface.”
In the datacom world, the path to market acceptance runs directly through Cisco Systems (www.cisco.com), the largest systems house in the space. It’s only a slight oversimplification to suggest that the evolution of datacom MSAs follows the ying/yang of attempts by industry to anticipate Cisco’s desires and Cisco’s judgment of the results. Nowhere is this clearer than in the case of the X2 and XPAK MSAs, two coincident and competing attempts to come up with a smaller version of the XENPAK module. Cisco chose the X2. As a result, “that platform will be a high-volume runner,” in the words of Tom Fawcett, product line manager for datacom at JDSU (www.jdsu.com). The XPAK volumes, concentrated in the server space, according to sources, will prove significantly smaller.
Cisco’s hand also shows in the market penetration-or, in certain areas, the lack of penetration-of the XFP, say sources. The XFP was offered as an attempt at telecom/datacom module convergence, anticipating that telecom systems suppliers would want something smaller than the 300-pin MSA and datacom suppliers (one in particular) would eventually want something smaller than the X2. However, Cisco has been slow to accept the XFP (the company did not provide a spokesman in time for this article), which has opened the door for another contender for a next-generation module, the SFP+.
As its name implies, the SFP+ is a slightly larger version of the SFP; it is also slightly smaller than the XFP. While specifications of the device remain in process, it is expected that the module’s increased size and generally more robust attributes versus the standard SFP will enable it to accommodate the electronics and power necessary to meet 10 Gbit/s requirements, particularly in the Ethernet and Fibre Channel arenas. (It will likely first appear as an 8 Gbit/s device for Fibre Channel, industry sources say.)
Module vendors are ready to embrace the new form factor, which doesn’t bode well for the XFP’s target of being a universal transceiver. “I would have to guess with the advent of SFP+ and the density that it enables, going forward we’re going to start to see a lot more Ethernet switches being designed for the emerging SFP+ standard, just because of the advantages that it provides,” predicts Rafik Ward, senior director of marketing at Finisar (www.finisar.com).
The fact that, according to some of the industry sources, Cisco appears to view the module favourably doesn’t hurt, either. “I think part of their rationale for looking at SFP+ is that they didn’t want to subsidise the XFP market-why should they let their competitors benefit?” comments Cornejo of the company he referred to not by name but as “the largest datacom player.” “So they’re actually making a pretty strong move to SFP+.”
Of course, Cisco isn’t the only company that uses transceivers, particularly for carrier networks. Yet even in telecom, one company can make a difference. Opnext is a member of the XFP-e MSA, a form factor aimed at a slightly larger version of the XFP that would accommodate some of the functions, tunability in particular, that currently fall outside the XFP’s capabilities. “That’s been mainly driven by a very large telecom customer; I think they’re ranked number one,” says Cornejo. “I’m not sure why you need pluggability for tunable, but that is kind of its intention.”
Despite the arrival of the SFP+, the XFP won’t go away soon, principally due to its success in the telecom space. “We definitely see XFP being the successor to 300-pin in the SONET space,” comments JDSU’s Fawcett.
Lars Friedrich, vice president, product line management at DWDM/CWDM systems vendor ADVA Optical Networking (www.advaoptical.com), says his company has embraced the XFP for line-side use and is very interested in the SFP+ for client-side interfaces, with pluggability a major reason. The use of pluggable transceivers enables telecom system vendors to develop “colourless” line cards, which can be populated with pluggable transceivers to meet customer requirements. Similarly, pluggable transceivers also allow the creation of “regenerator on a card” designs (as opposed to back-to-back line cards), as well as the economical addition of such features as protection paths.
“What’s so attractive about that is the incremental cost for the added functionality for the second route for protection or for the second interface for a regenerator comes at the lowest possible cost because you’re not paying for an additional card or additional stuff around it,” says Friedrich. “Fundamentally, the transceiver boils it down to just paying for the additional interface.”
With the success of the XFP in telecom, as well as some areas of datacom, and the expected acceptance of the SFP+, industry observers believe that the market finally may be converging-as far as new system designs are concerned, anyway-from seven modules to two. “My understanding is that in the datacom arena, probably the future form factor is SFP+ because of the port-density requirement,” offers Eddie Tsumura, vice president of engineering and marketing at ExceLight (www.excelight.com). “But in the SONET arena, for SFP+, the MSA doesn’t specify a SONET application. So my understanding is XFP will be used for a longer time in the SONET application.”
“The XFP is going to be around for at least 5 years, probably longer,” Inniss predicts. “And that’s because it really was just designed into the carrier equipment. It has some minimum lifespan; they’re not going to go back and redesign line cards just because there’s an SFP+.” The fact that the SFP+, at least in its initial incarnation, probably won’t be robust enough to support reaches beyond 10 km-and won’t get beyond 40 km anytime soon, according to the module vendors contacted for this article-also bodes well for the XFP’s lifespan in carrier equipment. Meanwhile, Friedrich expects to see thermally tunable XFPs at some point next year.
So perhaps there is hope for convergence, and fewer form factors, in the future. “We’re hopeful that the SFP+ is going to be a common ground for 8.5 and 10 Gbit in the volume portion of the enterprise,” concludes Fawcett. “And our belief is that the 300-pin market will continue to migrate over time towards XFP. If we can get that kind of clarity and that kind of limitation on platforms, I think we’ve done a lot to simplify 10 Gbit.”