A new industry standards group will tackle an issue that’s been a hot debate in communications for at least two years—should the next big leap for Ethernet be to 400 Gbits/second or a Terabit/s? The IEEE 802.3 Industry Connections Higher Speed Ethernet Consensus group aims within a year to pick a target and kick off a formal standards effort.
This comes as the IEEE 802.3 group overseeing Ethernet officially publishes its Ethernet Bandwidth Assessment report. The report, developed over the course of the last year, validated the group’s conclusions from an earlier study that demand for bandwidth is doubling about every 24 months. At that rate, networks will need to support capacity requirements of a terabit per second in 2015 and 10 terabit per second by 2020, it said.
“Everyone is worried about the coming tsunami of data,” said John D’Ambrosia, a veteran Ethernet standards organizer and chair of the new consensus group. “There’s a lot of nervousness out there about how people will do it in a cost effective manner,” he said.
The problem is there are no economically viable solutions on the horizon for terabit rates today, and bleeding-edge users say the more practical 400 Gbit/s target is not enough to meet their needs.
Engineers are currently hammering out specifications for 25 Gbit/s serial lanes and starting work on 50 Gbit/s channels at the Optical Internetworking Forum. It would take 20 to 40 of those fastest channels working in parallel to deliver TBit/s Ethernet rates.
The consensus group could decide to set both 400G and a Tbit as parallel goals. The last big Ethernet effort simultaneously delivered standards for 40 and 100G to serve different markets. The consensus group could also punt, putting out a call for interest for the next standard, but leaving the choice of a data rate to the follow-on group.
Whatever the decision, D’Ambrosia predicts within a year the consensus group will kick off a formal next-gen standards effort. He expects the follow on effort to take about three years, a year less than for typical next-gen Ethernet standards.
“It will be quicker this time because the bandwidth tsunami is coming–four years is too long, and I think people realize that,” he said.
D’Ambrosia expects collaboration between the IEEE and ITU to ensure that the optical networking standards set by the latter group are in harmony with the IEEE effort. Toward the end the consensus group will have its first meeting September 23 in Geneva on the heels of an ITU gathering.
Past experience have shown solutions using even 16 lanes are at the edge of what is economically practical. That’s in part because such dense parallel links push the limits of the number of pins that can be placed on a chip.
Thus some experts speculate the next big leap could require radical new modulation schemes or other breakthrough techniques. “We need a fundamental technical paradigm shift to move to terabit, and nothing has emerged at this point,” said D’Ambrosia.