SFPs challenging tunable lasers in metro
By Meghan Fuller
Tunable lasers have long promised dynamic reconfigurability and on-the-fly provisioning, but these applications have yet to materialize. In the interim, sparing remains the key driver for tunable-laser use in the metro. Yet, tunable lasers come with a considerable price tag, prompting carriers to consider small-form-factor-pluggable transceivers (SFPs) to perform the same function at a much lower price point.
The advantage of SFPs, contends Atikem Haile-Mariam, DWDM product-line manager at Finisar (Sunnyvale, CA), “is not so much for sparing as it is flexibility for adds, moves, and changes.” Carriers are using SFPs to mitigate the effects of increasingly unpredictable traffic patterns, he says. In the past, carriers would have to dedicate a whole line card to a fiber or pair of fibers in a given wavelength. To increase the speed of the wavelength or configure it for a different traffic pattern, carriers would have to remove the line card. In older systems, a line card couldn’t be removed without powering down, he adds, “so it was a big deal. The whole name of the game here is that you don’t want to mess with the line card.”
“There are sparing options with classic technologies, like buying a full card for $10,000. But sparing multiple wavelengths requires multiple cards,” notes Glenn Thurston, vice president of marketing for metro edge systems supplier BTI Photonic Systems (Ottawa, Ontario). “In the case of our customers, their sparing strategy is a plastic baggy of SFPs. It has totally changed the paradigm.”
Despite the ease of use provided by SFPs, tunable-laser manufacturers are not particularly concerned that the SFP will shrink their addressable market. For example, Julian Osinski, vice president of marketing at Santur (Fremont, CA), reports that the bulk of his company’s business remains in the long-haul market rather than the metro. Tunable lasers aren’t yet meant to compete with the lower-cost (i.e., fixed) lasers used in metro systems today, he says. Directly modulated WDM lasers sell for between $400 and $500, whereas a tunable laser might be priced at $1,000, plus $500 for the external modulator. “In general, I think tunable lasers aren’t penetrating very far yet in the metro,” he concedes.
“Having a pluggable transceiver certainly eases inventory and sparing, which has been one of the major selling points for tunable lasers over the last few years,” observes Kevin Affolter,vice president of marketing at Agility Communications (Santa Barbara, CA). “But SFPs do not solve the problem. The customers still have to manage, say 32 or 64 different wavelength codes. Which means that their vendors-the people who are providing the SFP transceivers-need to provide different laser chips that have different wavelengths, and they have to forecast manage all of that, which is still a pretty significant headache.”
“Because of the price pressure on pluggables, they will never have the same performance as you can get with a full tunable laser,” acknowledges Per Hanson, director of business development at ADVA Optical Networking (Munich, Germany, and Mawah, NJ). For this reason, he envisions two separate markets developing in parallel.
Pluggables will be more attractive for smaller networks of <150 km. But “in bigger networks where the higher-end performance of the optical signal is critical to meet distance needs, you will have tunable lasers,” explains Hanson. He notes that ADVA is “putting a lot of weight on using pluggables because we do see them as a very effective way to reduce costs in smaller systems in particular. And also because we advocate the use of CWDM and ability to grow into DWDM, and pluggables inherently support this as well.”
Finisar’s Haile-Mariam is more than happy to leave the long-haul and ultra-long-haul markets to the tunable-laser manufacturers. “The real battle is in the metro,” he asserts. “We think once DWDM SFP specs are firmed up and [modules] start shipping in quantity, these unlocked, cooled solutions will win out over any kind of tunable solution [in the metro].”
Today, 850-nm and CWDM SFPs are the norm, but several companies have released preliminary versions of DWDM SFPs over the last 18 months, including Finisar, LuminentOIC, and Eudyna Devices USA in conjunction with Fujitsu. Others remain interested in the product, including Agilent Technologies, JDS Uniphase, and Tyco Electronics. These vendors, along with Finisar and TriQuint Semiconductor, launched the DWDM SFP multisource agreement (MSA) in February 2004. TriQuint has left the optical-module business with the sale of its former Agere assets to CyOptics; according to CyOptics sales and marketing vice president Ali Abouzari, his company does not intend to market TriQuint’s old SFP product line and will instead leverage the InP semiconductor and related component technologies it gained from the acquisition.
Despite the number of vendors working to develop MSA-compliant DWDM SFPs, Haile-Mariam reports that the industry will not see heavy volumes until the first quarter of next year. “There is a huge pent-up demand,” he says, “No question about that. But the vendors are still catching up to that demand.”
Thurston also sees big demand for the ITU-T grid devices. “When I talk to SFP vendors, they are all building the low-cost 850-nm stuff and CWDM. They ask us, ‘Do you really have a need for DWDM?’ Our response back to them is, ‘If we had product that was running at a quality level, we could ship some today for DWDM systems.’ Our customer base is pulling pretty hard for this.”
In the interim, carriers continue to leverage the SFPs available today. Verizon currently specifies the use of pluggable SFP/XFP optics on the client side, since the wide variety of interface requirements is best served by pluggable technology, according to Stuart Elby, vice president of Verizon Network Services (New York City). “Moreover,” Elby says, “we can add pluggable modules as required by service demand to reduce up-front cost. On the network side, we strongly prefer transponders with tunable lasers at 10-gig rates since the premium is relatively small.”
Glenn Wellbrock, director of network technology development at MCI (Ashburn, VA), confirms that MCI also favors the use of SFPs on the client side. “SFPs eliminate up-front capital costs because you can buy a card unpopulated,” he says. “In the metro environment, a lot of connections are made intra-office. If you’re just going from box to box in the same central office, all you need is a very inexpensive pluggable SFP. Most of them have multiple ports per card, so you would just buy them as you need them.
“We do like tunables. We use them today and will continue to go that route but only in ultra-long-haul, the multiple hundreds of kilometers where there is usually only one interface per card. In other words, wherever we have a 300-pin MSA card, that’s where we believe it ought to be tunable.”
But what if you could make an SFP that was also tunable? Are tunability, pluggability, and small form factor mutually exclusive?
“If you could make a tunable SFP, that would completely solve your inventory problems,” says Agility’s Affolter. “And it would also not pigeonhole you into a system that couldn’t be reconfigured later on.”
Agility is working on a tunable SFP, according to Affolter, but such devices are still a couple of years away. Santur’s Osinski confirms that his company is also interested in developing a tunable SFP, though the addition of tunability presents formidable design challenges. Customers will not pay more than a 5% premium for a tunable SFP versus a fixed SFP, he says, so hitting the price point is the first hurdle. Manufacturers also face the daunting task of meeting the requisite reach and output power requirements in the smaller form factor.
For this reason, both Affolter and Osinski believe tunability may be more readily achievable in a larger package at higher speeds. “At 10 Gbits/sec, the feasibility of having a tunable XFP or tunable transceiver is probably better than it is at 2.5 Gbits/sec with an SFP, because the size and power constraints [at 2.5 Gbits/sec] are that much greater,” explains Affolter. In addition, 10-Gbit/sec optics tend to extract higher prices.
Wellbrock confirms that MCI would be very interested in a tunable XFP, calling it “a win-win situation. Priority-wise, a pluggable SFP with a fixed laser is better than a fixed MSA, but it’s not as good as a tunable SFP.” However, assuming price points are the same, the tunable 300-pin MSA might still be the best option. “In an MSA, you have the additional space and resources available in the can so you wouldn’t be as constrained as you would be in an XFP,” he says.
SFPs: Beyond Sparing
The reconfigurable optical add/drop-multiplexer (ROADM) market is one of the hottest in the industry, with carriers like Verizon and SBC issuing RFPs that specify ROADM capabilities. Will small-form-factor-pluggables (SFPs) emerge as a viable alternative to tunable lasers in this market as well? “There’s a much more complicated SFP story that involves how you add, drop, and manage wavelengths,” argues Atikem Haile-Mariam, DWDM product-line manager at Finisar (Sunnyvale, CA).
The first generation of ROADMs featured a stable, locked laser source that would enable the ROADM to direct certain wavelengths in certain directions. Next generation ROADMs, by contrast, are coupled with a tunable-laser source to provide dynamic reconfiguration and therefore solve a lot of the bandwidth issues facing today’s carriers.
“Another way to solve [the problem] is by having very-low-cost DWDM SFPs at these remote sites or communications closets instead of a centrally controlled ROADM/tunable-laser environment,” says Haile-Mariam. If a neighborhood or business district suddenly requires more bandwidth or a different wavelength, the carrier can simply switch out SFPs or add more SFPs at that location, he explains.
The ROADM market is experiencing a lot of momentum, he adds, “and the tunable-laser guys want to hook into that ROADM equation. But after 18 months, it’s entirely possible that a narrowly tunable SFP will supplant a number of these applications and work with a ROADM or with the traditional point-to-point architectures that exist now.”
Glenn Wellbrock, director of network technology development at MCI (Ashburn, VA), agrees that tunable SFPs hold a great deal of promise, assuming they meet power and reach requirements. “We’re continuing to push them,” he says. “There is no reason why we wouldn’t want everything pluggable if you could really do it. I think there’s very little reason not to.”