"10-Gbit/s Ethernet combined with IP really accelerates the demise of Sonet networks," Stitt said during his keynote speech at yesterday's Light Reading Live! event, "10-Gbit/s Ethernet Components: Ready for Prime Time." The day-long session, moderated by Heavy Reading analyst Simon Stanley, focused on the semiconductors, optical modules, and network processors being targeted at 10-Gbit/s Ethernet.
Stitt had predicted the death of Sonet in 2000 at George Gilder's Telecosm conference, raising some protests from the non-believers. Most 2000 predictions are about as valued as Gigli DVDs, but Stitt's seems as if it's going to hold up.
Stitt emphasized that it's going to take a long time before the U.S. installed base makes the shift to Ethernet, but he thinks Ethernet has won the long-term race. What's going to trigger the change? Stitt said he heard the answer over lunch, when a banking CIO described how his industry is getting weaned off mainframe computers: "It's not really the cost. All the people who know how to run these mainframes are dying."
More literally, they were retiring, Stitt noted (sorry, we don't live in an Agatha Christie novel), but his point was that Sonet, like mainframes, has a knowledge pool that will be shrinking. The new talent in networking is "all going to be well versed in Ethernet," he said.
Stitt underscored that the 10-Gbit/s generation differs from previous Ethernet generations because telecom has taken a stake this time. Rather than accepting another best-effort technology, customers are expecting 10-Gbit/s Ethernet to come with more intelligence for features such as security and quality of service (QOS).
"Network managers won't accept the bandwidth without the control," he said. "This will be a very different market than the 90s, where bandwidth was everything," Stitt said.
Stanley emphasized that point in a later session, noting that Gigabit Ethernet had been presented to the world as a "dumb pipe" that happened to be really fast. The same tactic won't work this time. "We've now moved on to 10 Gbit/s, and QOS is important because of the new applications," he said.
One trait that didn't change from previous generations was the rapid decline in prices. Announced 10-Gbit/s Ethernet prices have hit $5,000 per port, compared with more than $50,000 at the end of 2002 (see Cisco Bombs 10-GigE Pricing). The sinking price was what pushed the technology out of the lab and into production networks during the past year, Stitt said.
Not everything is sunshine and rainbows with 10-Gbit/s Ethernet, though. Stitt decried the number of module options out there, with Xenpak, XPAK, X2, and XFP all vying for attention. "Frankly, our customers get grumpy about that as well. They don't want to have to keep a lot of spares around. Every time we go to a new low-cost form factor, my phone starts ringing," Stitt said.
It's also worth pointing out that a lot has happened to Sonet since Extreme launched its attack on it four years ago (see Extreme Launches A Sonet Killer). Next-generation Sonet/SDH developments have enabled carriers to roll out Ethernet services over their existing infrastructures, and the cost of doing this has been slashed, thanks to the arrival of off-the-shelf Sonet/SDH silicon.
In a recent Light Reading poll taken by nearly 300 readers, close to 50 percent of respondents thought Sonet/SDH had changed more than Ethernet since the year 2000. The technologies tied, each getting 37 percent of the vote, on how much impact they'd made on carrier infrastructure in that time period (see Why Sonet Chips Are Sexy).
— Craig Matsumoto, Senior Editor, Light Reading
For more on this this topic, check out:
- The Heavy Reading report:
— 10 Gbit/s Ethernet Components: A Heavy Reading Competitive Analysis
For further education, visit the archives of related Light Reading Webinars:
For a schedule of future Light Reading Live! events, click here (.pdf format).