100GbE per Lambda for 400GbE: The Work has Just Begun
Ethernet Alliance chair John D'Ambrosia unilaterally cancels the networking industry's vacation plans. It's time to work on 100GbE per lambda and 400GbE.
Within the enterprise, many organizations are still working on migrating from 1GbE to 10GbE. In the world of industry groups like the Ethernet Alliance and standards bodies like the IEEE, however, the focus is on the development and standardization of technologies far more advanced. Next week in San Jose, California, the Ethernet Alliance and OIDA will host the 100GbE per Lambda for Data Center Workshop, focusing on bringing single-wavelength 100G Ethernet to data centers to enable the 400GbE future.
The workshop grew out of the Ethernet Alliance TEF 2013 and took on more significance after the IEEE's 400GbE Task Force May Interim meeting, as John D'Ambrosia, chair of both the Ethernet Alliance and the IEEE P802.3bs 400GbE Task Force, told me. The 400GbE Task Force's timeline requires that all new proposals be received by November of 2014 and drove significant interest in 100GbE per lambda for 400GbE.
"At the 400GbE Task Force meeting, there was a straw poll on how to make 400G happen in the future. The choice for 2 km and 10 km single-mode fiber solutions, the solutions that got the most amount of votes, were assuming 100G effective bit rate to wavelength. It was assumed that we would achieve 400G through some multiplexing-type solution," D'Ambrosia said.
The solution that won the most support in the nonbinding straw poll was for each wavelength to have an effective bit rate of 100G: a 4 x 100 solution. Now the challenge becomes making 100G economically feasible for the data center environments that will use it.
"Look at what's happening in the data center and networking today. We see 100G being deployed at the edge and 40G being deployed in the data centers. Why? Cost. They have different cost models. So how can we make 100G on a single wavelength cost-effective for the data center environment?" D'Ambrosia asked.
Work has just begun on the problem, D'Ambrosia happily admitted. A new spec isn't going to emerge from a day-and-a-half workshop. But what he hopes will happen, and what sessions on topics like the drivers and applications for 100GbE per lambda, the 100GbE per lambda physical layer, and what makes 100GbE per lambda a compelling investment will help accomplish, is discussion. Industry leaders must come together to decide what kinds of investments need to be made and what conditions the technological development will require from—and impose on—the industry.
Bandwidth needs continue to grow in the data center, driven in large part by applications like streaming video. This doesn't look likely to slow down soon, especially given the development of new applications—and entire new classes of IP-connected devices such as those promised by the Internet of Things—coming into the market. Bandwidth and connectivity create the environment that such devices and applications need to succeed, a sentiment with which D'Ambrosia enthusiastically agrees.
Automotive Ethernet is one of his go-to examples of what could happen.
"I'm terrified of that one. It's going to be moving all around the network. That's going to be a distributed problem based on where traffic is. And when you think about the implications within wireless backhaul networks, you have to support that traffic demand wherever it's needed, not just whenever, at a given location," D'Ambrosia said. Concepts like automotive Ethernet and any of a number of other new technologies depend on bandwidth and connectivity to succeed.
"If the bandwidth is there to provide a great user experience, the applications are successful and the bandwidth demand goes up. If the bandwidth isn't there, it hinders the applications," he added.
That's where efforts like the 400GbE Task Force and the 100GbE per Lambda for Data Center Workshop come in. To support and advance an entire ecosystem of new technologies, industry leaders must come together to make the necessary backbone affordable and feasible for the data centers that need it—and they must do it soon.
"Given the adoption of the timeline for 400G, I've canceled industry vacation plans for this year," D'Ambrosia laughed.
Header photo courtesy of Shutterstock.
Jude Chao is managing editor of Enterprise Networking Planet. Follow her on Twitter @judechao.