OSA 2011: Optical World Faces Hipster Challenge
The topic provided some of the most interesting insights during panel sessions at Monday's Executive Forum, which as usual served as an OFC/NFOEC preview.
Google's case was particularly spotlighted, partly because the company landed two panel spots (both filled by Senior Network Architect Bikash Koley, due to a colleague's illness). Koley made it clear that he thinks the traditional telecom cycles are too slow for Google and don't produce the right kinds of products anyway.
That's because standards too often aren't developed with the input of the ultimate buyers, he said. By the time products arrive, they're too big and/or power-hungry to suit the next-generation needs that the user is building for. Koley stepped through an example of how it all goes wrong, drawn not from Google but from his experience at a large equipment vendor. (He didn't specify which one, but his resume includes time at Ciena Corp. (NYSE: CIEN))
The problem stems from each vendor taking its cues from a neighboring step in the supply chain, rather than going to the source. The result, in Koley's example: A product that arrived years late. "Not talking to a user gives you the wrong time horizon," Koley said.
One alternative is to bypass the standards bodies and develop a multisource agreement (MSA), a tactic that's worked for transceiver modules.
"Any time there's a void and there's enough need, especially if there's enough bandwidth that needs to be deployed, there's a vehicle" thanks to MSAs, said Donn Lee, a senior network engineer at Facebook. (Lee appeared on a panel separate from either of Koley's.)
Need evidence of how this might work? Koley pointed to the recently ratified 10x10 multisource agreement (MSA), which was created with input from a spectrum of companies including not just module suppliers, but also cable operators, telcos, Ethernet service providers and Google.
"We informally sat down in a room for a whole day," Koley said. "It was very informal. It was not a standards body. It was brainstorming how to solve the problem and what we need from the industry. Turns out our needs were pretty similar. We do not talk enough."
Towards the end of the day, Alain Couder, CEO of Oclaro Inc. (Nasdaq: OCLR), summed up the situation concisely: "We have a totally different mindset in the new set of users that is going to force the telecom vendors to totally evolve in the way they work with optical component vendors."
Something else that rankles the data center crowd is the level of scrutiny telcos place on optical products. This came up in October, when Google and Finisar Corp. (Nasdaq: FNSR) suggested that the industry let some devices meet relaxed standards. (See Google Loosens Up on Data Center Optics.)
Google also doesn't like telco features such as Optical Transport Network (OTN) support. "We don't need it; we won't pay for it," Koley said. But on stage, he didn't try to bully equipment vendors into creating non-OTN versions of their gear. Not exactly, anyway; he said he'd like to see modular development, so that Google can omit unwanted OTN or GMPLS software.
OTN is a good example of the rift between telcos and a newer breed of network owner that includes not only Google but Comcast Corp. (Nasdaq: CMCSA, CMCSK) as well. The latter camp doesn't have an old installed base of TDM services flitting about its packet network.
For many years, the telecom industry has talked about migrating those TDM services into packet form, but it's just not happening. And that means content providers will be wrangling with TDM-oriented features for quite some time.
"I can't remember the last time we retired a service, and that may be part of the problem -- we're afraid to retire a service," said Scott Mountford, principal network planning engineer at AT&T Inc. (NYSE: T).
— Craig Matsumoto, West Coast Editor, Light Reading