Conferences such as OFC/NFOEC are, of course, great places to pick up news of technological breakthroughs, hints of application trends, gossip, and the odd but interesting factoid. I've written about many of the tidbits I encountered in Anaheim already. But here are a few that haven't yet escaped my notebook or that could use some additional color.
The theoretical floor to the current decline in cost per Gbps is about $5.00. That's where the graph LightCounting's Dale Murphy displayed at a dinner the market research firm hosted March 20 leveled off. I even asked him if I had reached the correct conclusion, barring any significant change in fundamental technologies. "Yup, I guess so," he replied.
Silicon photonics promises to revolutionize photonic integration. But it's not the only viable means of photonic integration. The current generation of silicon photonics devices hints at what the future may hold. But the approach won't sweep away other strategies until volumes reach numbers high enough for its promised cost savings to kick in, and at least some of those devices can support reaches encountered in the outside plant. (Yes, Kotura's 100-Gbps QSFP is a start in this direction. But that only covers 40 km right now.) Until then, approaches based on indium phosphide look equally viable.
That said, there should be plenty of applications in the data center for this early generation of silicon photonics. Such integration likely will be the only way optical technology reaches the price, power, and footprint thresholds necessary to make a real dent in intra-rack applications. Silicon photonics surely will get a boost when Intel's efforts move from engineering samples to actual products. The architecture Intel and Facebook unveiled last January at the Open Compute Summit gives silicon-photonic-device developers something to shoot for. Let's see if they hit the target.
If you want decent Internet access in Anaheim, don't stay at a hotel with a façade shaped like a castle. This probably goes as well for façades shaped like pirate ships, space ships, or cartoon characters. Or any hotel in which all the rooms are accessed via outdoor walkways that face out on a courtyard.
CPAK will provide a major test of Cisco's market influence. Cisco used to be the 500-lb gorilla of the datacom world. The company pretty much could dictate transceiver form factors and features, for better or worse. However, that influence has appeared to wane in recent years with the rise of mega data-center users such as Google and Facebook, whom vendors perceive as a market niche on their own.
Cisco's decision to go in its own direction for 100-Gbps modules with CPAK will test whether the company still can throw its weight around. One intended byproduct of the CPAK effort was to create competitive advantage with a module Cisco deemed superior to the CFP and CFP2, while driving up the prices of those devices for its competitors; removing Cisco's volumes from the sales of CFP2 modules theoretically should slow their descent down the volume-driven price reduction curve. If this strategy works, Cisco still has the old market-shaper magic. If not, well…
Past Lightwave Issues