Ethernet Group Starts to Talk Terabit

The Institute of Electrical and Electronics Engineers Inc. (IEEE) is taking the first step towards standardizing the next speed of Ethernet, with one key question still being whether the goal should be 400Gbit/s or 1Tbit/s.

The standards body has created the Higher Speed Ethernet Consensus group, to be chaired by John D'Ambrosia, chief Ethernet evangelist at Dell Technologies (Nasdaq: DELL), who's been a driving force for high-speed Ethernet, from 10Gbit/s through the current effort. (Note how the new speed has to be higher.)

D'Ambrosia's hope is that this ad hoc group will formulate a call for interest (CFI) -- the first step in creating an IEEE standard.

Why not just float a CFI right now, rather than talk about it? Because it's worth it to make sure that any proposal has a good shot at getting the 50 percent approval a CFI requires, D'Ambrosia says.

He also wants as many opinions as possible to be fed into the process. That didn't happen with 100Gbit/s Ethernet, which got delayed when the server community started pushing for 40Gbit/s.

"When you look at that project from when I held the first call -- not necessarily as an IEEE project but the first call to start that -- that was December 2005. The spec was published in June 2010," D'Ambrosia says.

D'Ambrosia actually started the higher-speed Ethernet process last year by working on a bandwidth assessment for the IEEE -- making sure, before starting any standard, that the group had some understanding of future bandwith requirements. The study, completed in July, found that 1Tbit/s links were likely to be needed in 2015, and 10Tbit/s by 2020.

That doesn't imply that the IEEE has to crank out a 1Tbit/s standard in three years, D'Ambrosia says -- ten 100Gbit/s lines could be used instead. The point is just that the IEEE now has a common understanding of the problem.

"The doubling of bandwidth every 18 months still seems to be a reasonable approximation for where bandwidth is going," D'Ambrosia says.

Some obvious issues remain, such as whether the next standard should be 400Gbit/s or 1Tbit/s. The latter is going to be needed, but there's no clear way to make it affordable.

"During the course of the last speed, there was a lot of 'Build it, and it will sell.' Then there was sticker shock," D'Ambrosia says.

There's also a matter of scope to consider -- that is, exactly which specifications the standard should tackle. It's not as glamorous as saying "terabit," but it's an important factor.

— Craig Matsumoto, Managing Editor, Light Reading