SANTA CLARA, Calif. — Separate efforts to drive tomorrow's Ethernet networks faster and to get today's more widely deployed were on display at the Ethernet Summit here. Speakers also touched on the lack of standards in cloud computing and the still evolving nature of software-defined networks.
An initial standard for 400 Gbit/second Ethernet should be finished by 2017, hopefully laying a foundation for work on a Terabit spec, said John D'Ambrosia, a veteran of many IEEE 802.3 efforts. Mega data centers, Internet exchanges and wireless and cable providers demand the faster data rates, he said.
"Terabit Ethernet got delayed but it's on the horizon," D'Ambrosia said, showing a likely road map of high-speed Ethernet efforts (above).
Work on the 400GE standard officially starts May 12 at a meeting in Norfolk, Virginia, with this year focused on architectural proposals, he said. It will most likely begin with concepts for 16 lanes of the 25G serial links just now emerging from the lab. Ultimately it will need to cover a wide array of chip, board and short- and long-haul system links.
D'Ambrosia said the so-called CDFP 16 x 25G connector (below) is a practical starting point. However, "it's not necessarily the lowest power or cost solution -- ultimately it’s a transition technology," he said.
Engineers have started the hard work of developing 40, 50 and even 100G serial links, including a June workshop on sending 100G over a single optical wavelength, he said. Difficult as the task seems at the moment, "the most direct path to 400GE and Terabit Ethernet is getting to 100G serial links -- that’s where we need to be," he added.
"I love this industry because we do rise to the occasion, but let's not kid ourselves we have problems -- there are some pesky laws of physics to deal with," he said.
Next page: 400GE up and running