Systems vendors see no time to wait for 100 Gbps
JUNE 1, 2009 By Stephen Hardy -- The OIF has embarked on a well-documented campaign to create a component and subsystem ecosystem for long-haul and metro 100-Gbps systems. However, the effort may not be moving fast enough to suit the market.
JUNE 1, 2009 By Stephen Hardy -- The Optical Internetworking Forum (OIF; search Lightwave for the OIF) has embarked on a well-documented campaign to create a component and subsystem ecosystem for long-haul and metro 100-Gbps systems. The goal is to obviate the need for proprietary system designs that don't leverage the potential industry-wide cost savings that merchant chips and modules can provide, as well as create a better chance for 100-Gbps interoperability.
However, as companies such as ADVA Optical Networking, Ciena, and Nortel have demonstrated, the effort may not be moving fast enough to suit the market. (Search Lightwave for ADVA, Ciena, and Nortel.)
The OIF has attempted to concentrate industry efforts in one spot: dual-polarized quadrature phase-shift keying (DP-QPSK) as a modulation format, with coherent detection as a boost. The group has done an admirable job of developing technological consensus among companies that commonly operate alone; there are plenty of vendors working on pieces of the DP-QPSK/coherent detection puzzle.
The problem, say sources at the systems houses, is that this puzzle will be very difficult to complete. Meanwhile, they have customers who want 100-Gbps now or in the near future.
Perhaps the OIF can take some solace in the fact that Nortel and Ciena have produced variants of the OIF's chosen technology. Nortel uses two 50-Gbps DP-QPSK subcarriers spaced 20 GHz apart to create a composite 100-Gbps link. Ciena (whose vice president of network architecture, Joe Berthold, is a frequent spokesman for the OIF) has chosen DP-DQPSK with direct detection for the 100-Gbps capabilities now offered with the CN 4200 RS FlexSelect that are about to see their first deployment.
The Ciena approach can be seen as a steppingstone to the OIF's eventual goal, Berthold says. DQPSK is already in use for 40-Gbps networks. The distances Ciena plans to support probably won't require coherent detection, Berthold continues, adding that coherent detection also will require some "pretty sophisticated" ASICs for signal processing and advanced analog-to-digital converters that are still under development. Even with direct detection, Ciena had to develop high-speed ASICs internally to make the system work.
The company publicly demonstrated the technology for the first time at Supercomputing 2008 last fall. Berthold said that customer feedback from that demonstration convinced Ciena that it should commercialize its approach now, rather than wait for the development of off-the-shelf building blocks.
A lack of readily available components also led ADVA Optical Networking to go its own way at 100 Gbps, according to Jim Theodoras, the company's director, technical marketing. With a focus on 100-Gigabit Ethernet metro and regional applications, ADVA has announced the pursuit of differential phase-shift keying, three-level amplitude-shift keying (DPSK-3ASK) as a modulation format.
Again, it's the ICs that Theodoras cites as the sticking point. "These chips are very big, they're very complex, and it's going to take a while for these things to make sense," he says of 100-Gigabit Ethernet devices. "If you're doing a 10-gig PHY, where the volumes are in the tens of millions, it's not hard to justify another mask set. But when you're talking about initial 100-GigE volumes, which are very low, it may justify a development effort, but does it justify a new custom chip or justify a new mask set? And often, that's a tough business case to sell."
Without the chips, systems houses are in a tough spot, Theodoras asserts. "The fact that we've gone down this other path says we see a reason to do it. It wasn't just because we like to play with new modulation formats. We see a demand for 100-GigE now," he explains.
"If the standards bodies, the ITU or anyone, adopts other formats as standard we're obviously going to go down that path too. But we need something in the interim that can meet this bandwidth need that we're seeing without needing to wait for this type of industry investment to occur," he concludes.
Not surprisingly, module vendors aren't too happy to see their potential customers striking off on their own paths. And just because a systems supplier hasn't announced a 100-Gbps offering doesn't mean they aren't working on one, sources at the module makers report.
For example, several companies are working on proprietary algorithms to combat impairments such as chromatic and polarization-mode dispersion, says Mintera Vice President of Product Marketing Niall Robinson (search Lightwave for Mintera). "Those companies really want those algorithms embedded in ASICs -- and they want to buy modules with those ASICs embedded in them," he says. "We cannot build a module and sell it to only one company. That's not a good business case."
In some ways, the module vendors find themselves in the same boat as their potential customers, since they also need silicon for their products. "A lot of folks are hoping to look at the transmitter as a merchant silicon path. But not everyone is comfortable with that because I think the magnitude of the receiver problem is more significant than the transmitter, but the transmitter is still not a walk in the park," says Matt Traverso, senior technical marketing manager at Opnext Inc (search Lightwave for Opnext).
Thus, it increasingly appears that while the OIF may succeed in having most of the industry on the same 100-Gbps page when its work is completed, that page won't be the only one in print.