Cisco Starts Picking XFP

Cisco Systems Inc. (Nasdaq: CSCO) has picked at least one vendor that could provide it with 10-Gbit/s Ethernet optical modules that conform to the XFP design. The lucky winner is reportedly startup Ignis Optics, which was acquired recently by Bookham Technology plc (Nasdaq: BKHM; London: BHM) (see Bookham Buys Some More and Bookham Completes Ignis Buy).

Ignis did not get a design win with Cisco, but industry sources say its XFP module has been qualified for use, meaning Cisco engineers have tested the part and are confident it will work.

Cisco is likely to qualify multiple XFP vendors eventually, so Bookham/Ignis will still have to battle to secure design wins. Cisco officials won't elaborate on which products have been qualified or when it might begin shipping boxes with XFP interfaces.

XFP is one of four multisource agreements (MSAs) for 10-Gbit/s Ethernet modules floating around the industry. The other three -- Xenpak, X2, and XPAK -- split the 10-Gbit/s stream into multiple lanes, but XFP is a serial connection. That means it's more compact, but it's also a higher-end device, likely to start appearing later than the other three. Since Xenpak is only just getting going, it will likely be some time before any significant XFP rollouts begin.

Ignis was determined from the start to get in early. It chose to skip the other MSAs and go straight to XFP, and, by November 2002, the company claimed it was the only vendor showing live XFP demos.

Cisco has always been interested in XFP. "There have been groups working in the company on XFP since probably before the MSA [got started]," says Bruce Tolley, senior manager of emerging technologies. He won't disclose Cisco's schedule for ramping XFP, and he says availability of modules won't affect the timetable. He did note that Cisco has begun shipping gear containing Xenpak modules.

Since acquiring the optical components division of Nortel Networks Corp. (NYSE/Toronto: NT), Bookham has been busily snapping up smaller morsels such as Ignis (see Bookham Buys Nortel's Components Biz, Bookham Gets Thin Film Filter Bargain, and Bookham Gets a New Focus).

These deals are helping Bookham cement its standing with some large OEMs, says SoundView Technology Group analyst Dennis Gallagher. The resulting business is worth the acquisition outlay. "Just what they're able to push to a Huawei or a Nortel justifies these small acquisitions," he says.

While the Cisco qualification doesn't guarantee tons of business for Bookham, the firm now has an early "in." Still, Cisco has about a dozen more candidates that could also gain qualification.

"We are in the mix and expect to be on the list," writes Mark Sobey, senior vice president of marketing at JDS Uniphase Corp. (Nasdaq: JDSU; Toronto: JDU), in an email. Another likely pick could be Finisar Corp. (Nasdaq: FNSR), which won Cisco's "Supplier of the Year" award recently; like Ignis, Finisar is skipping straight to XFP and ignoring the other "X" MSAs (see XFP Gets the Fast Track).

Other XFP vendors are concentrating on non-Ethernet markets, a strategy made possible by the XFP module's ability to be used with any protocol. E2O Communications Inc., for example, claims to have taken XFP orders from one of the major Fibre Channel switch vendors. That probably means Brocade Communications Systems Inc. (Nasdaq: BRCD) or McData Corp. (Nasdaq: MCDTA).

— Craig Matsumoto, Senior Editor, Light Reading

waverunner 12/4/2012 | 11:19:41 PM
re: Cisco Starts Picking XFP The Cisco way is paved with gold. Why even bother with MSA's, standards committees. Lesson to young product managers: Ask Cisco.

sigint 12/4/2012 | 11:19:40 PM
re: Cisco Starts Picking XFP Well, there are good reasons for *anyone* to opt for the XFP, and I guess Cisco is no exception.

The XFP would likely be cheaper - the burden of serialization being in the device driving it and none of the lane de-skew requirements. The XFP is much smaller, which translates into more ports per line-card.

The only problem is the very high signalling rate. On plane Vanilla FR4, that would require careful design, but that's no rocket science.

However, the market could sway in Xenpak's favour, should someone come up with a reasonable copper-module to work off the XAUI interface. It might not be feasible to fit a copper phy in the XFP form factor, at least for the moment. A copper/fiber line-card port makes more sense than a separate line card to a market that abhors inventory.

deckchairs 12/4/2012 | 11:19:40 PM
re: Cisco Starts Picking XFP Waverunner speaks reality. Items to ponder:

1) Cisco has a knack for project what services providers WILL want, not what is expected/available now. The account teams will sell the Cisco vision, with attractive pricing incentives, as a BETTER alternative to whatever anyone else is hawking. Also, a large role-out generally takes over six months once the contrace is signed - A lot of technology shift can happen during that period.

2) Cisco has the guns to drive the IETF "standards" to its benefit over the long term. As for the ITU, I think OSI and H.323 are two examples of that body's impact. By the time the ITU "recommendation" (read: if you're going to implement this protocol, here's how we recommend you do it.)

3) Cisco is such a force in the industry that other vendors will see sales opportunities by conforming to whatever Cisco chooses to deploy. The Cisco powers have been very cautious in crossing the standard body picket lines, but they will do it if it is the best interest of Cisco.

SIVROCX 12/4/2012 | 11:19:36 PM
re: Cisco Starts Picking XFP I think that Ignis has it together and Cisco made a good choice. They, Ignis, have good backgrounds from Motorola, H-P, Tyco and who knows where else. They have Steve Joiner who can use his powers of persuasion to work the standards groups. I donGÇÖt know anything about their design but I am sure they have paid attention to detail when it comes to the things everyone else adds as an after thought like EMI/RFI and manufacturability. The only downside I see to all of this is working with Cisco. Cisco always wants you to give your stuff to them for less than it cost claiming that you are the most costly component manufacturer in the world and if you want to be one of their suppliers you will give it to them at the cost they request. Oh and a little gift for those in purchasing wouldnGÇÖt hurt as well. Good luck Ignis, hope you donGÇÖt loose your shirt! +£
Mr. Mutt 12/4/2012 | 11:19:36 PM
re: Cisco Starts Picking XFP The problem with a startup latching on to Cisco is that it can completely drain your company resources with no revenue for months or years; especially on emerging technology. Cisco design from product commitment to FCS are anywhere from 12-36 months. That is 1-3 years of drain on a company that cannot afford it.

HOWEVER...now that Ignis is not a VC funded, revenue hungry startup anymore, this shouldn't be a problem for them. But I am unaware of Bookham's AVL position within Cisco. Optics is a VERY heavily controlled commodity within Cisco and not easily penetrated. In addition, Cisco is doing vendor reductions with the goal of giving 20% of their suppliers 80-90% of the business. Will they opt to engage with a new company for these parts? Only if they offer something no one else can.
Ben Crosby 12/4/2012 | 11:19:35 PM
re: Cisco Starts Picking XFP >>>
However, the market could sway in Xenpak's favour, should someone come up with a reasonable copper-module to work off the XAUI interface. It might not be feasible to fit a copper phy in the XFP form factor, at least for the moment. A copper/fiber line-card port makes more sense than a separate line card to a market that abhors inventory.

I really doubt this would be the reason the market "swayed" in xenpak's favour - for a number of reasons.

1. at 10Gb/s rate, it is exceedingly questionable that anyone would want a copper interface - what would the distance limitation be - 3 meters ? - I confess, I haven't read the spec if one exists.

2. SFP already provides a common packaging for both fibre, and copper (1GbE copper modules do exist, and I think are limited to 30m distance).

Where I'm going with this particular point is that in a largely similar sized device at lower speeds, we have copper / fibre interchangeability today. It is possible to mix and match Copper and Fibre on a single Nx1Gb board. XFP is slightly larger than SFP - and I'm sure if the demand surfaces for 10G Copper, someone will deliver an XFP module to provide it.

3. XFP has already been selected by other Vendors. Cisco is hardly taking a major risk, okay, I know a lot of the vendors are startups, but the technology does work.

waverunner 12/4/2012 | 11:19:30 PM
re: Cisco Starts Picking XFP Mr Mutt,

The typical design cycle is anywhere between 8-18 months whoever you work with, you could work with those you know will have revenue or you could work with those you think will have revenue.

Many prefer the former, and so do those who have the cash to acquire. Now if you are seeking an excuse for not making that call, well that's different.

Obsolete 12/4/2012 | 11:19:28 PM
re: Cisco Starts Picking XFP Ben,

The IEEE 802.3ak Task Force is in the final stages of writing a standard for 10GBASE-CX4, a 10Gb/s copper interface with a slight modification to the XAUI interface to go up to 15m over an 8 pair twinax cable.

Some companies have also built XFP electrical cable assemblies that work over a similar range.
Ben Crosby 12/4/2012 | 11:19:27 PM
re: Cisco Starts Picking XFP Thanks for the info. 15m might be useful in a few applications.

Mr. Mutt 12/4/2012 | 11:19:19 PM
re: Cisco Starts Picking XFP I don't want to get into a pissing contest, but I personally know of two programs on significant platforms that are in excess of your perscribed 18 months design timeline today. Things may be different for optics, but at 10G I would suspect optics decision are made further out than most other components other than ASIC. Cisco plans out things years in advance, and my point was that if you are engaged with them from the outset it can be significant time before you actually see R on your I.
Sign In