back to article Forget terabit Ethernet, the next step is 400 gig – if we can afford the R&D

The current top-end Ethernet standard may be 100 gigabits per second, but don't expect the next step up to be one terabit per second. The days of Ethernet speeds improving by an order of magnitude are gone. Why? Because nobody wants to pay for the necessary research and development – nobody in the US or the EU, that is. "My …


This topic is closed for new posts.
  1. Anonymous Coward
    Anonymous Coward

    10-bit multiplexed optical

    Just multiplex 10-bit (10 colour) optical gigabit ethernet.

    1. Anonymous Coward
      Anonymous Coward

      Re: 10-bit multiplexed optical

      Uh, you do realize that what you propose would provide 10 gigabit ethernet, or 100x short of what they're trying to achieve?

      What I think you meant to suggest is possible, but it would be more expensive and bulky than using 10 separate 10G connections. The fiber is cheap, it is the equipment on either end that costs money, and being able to cram 10 lasers (well 40, since 100G ethernet uses 4 lasers over 4 individual fibers already) into a small space isn't going to be cheap, nor would separating out all those wavelengths at the other end.

      Your idea is like someone saying, "I need a computer 4x faster than the one on my desk" and you say "no problem, we can stack four computers on top of each other!"

      1. Nigel 11

        Re: 10-bit multiplexed optical

        Optical wavelength demultiplexing ought to be pretty straightforward. You use a prism or a grating. I'm not an engineer but I suspect the demultiplexer is the least of the problems!

        The problem is that the same mechanism that separates wavelengths in the prism causes dispersion of signals in the fiber. At best the various optical carriers will get time-skewed in transit and that has to be compensated for. That's something that will vary with the temperature of the fibre.

        At worst, dispersion in the fibre will scramble the data modulated onto the optical carriers. If that's not a killer you have to make modulators and a multiplexor and a coupler. Rather them than me.

        It's far easier to transmit 40 packets in parallel down 40 fibres, than to try to make a 400Gbps single channel. Which makes me wonder, why is a 400Gbps single channel needed at all? c.f. one 400GHz CPU -- is that physicaly possible -- compared to the 1000 400MHz cores that you might find in a GPGPU.

        1. Anonymous Coward
          Anonymous Coward

          Re: 10-bit multiplexed optical

          If when you had wanted to upgrade from fast ethernet to gigabit, they'd taken a shortcut and used cables composed of 80 strands of twisted pair instead of sticking with the standard 8 strand cat5 type cable, would you have been as willing to do it?

          Multiplexing wavelengths isn't impossible, but it isn't so easy or cheap in a small form factor like you'd want to be able to use to plug your fiber into. If the place you plug it in takes up a 4" wide section of 4U high you wouldn't get much rack density.

  2. PacketPusher


    You mean like 640KB is all the memory we will ever need?

  3. earl grey

    the money is there

    It's just all being funneled into executive salaries for a bunch of greedy whinger who complain that users are "filling the pipes" and wonder even roll out the current levels of fibre to everyone. It's not they can't afford it; it's a conscious choice to screw the customer base.

  4. Anonymous Coward
    Anonymous Coward

    "China, on the other hand, is shoveling money into R&D, he said. "It's good news, bad news. The good news is there is some of the advanced work being done, and the bad news is that the jobs will probably shift over time to Asia."

    Not so fast! Your next article reveals that the air quality in these Asian countries is so bad that servers in data centers are being destroyed. It goes without saying that the increase in consumption of resources will only compound said problem, and the effect can only get worse. Maybe those job won't migrate to Asia if the tech won't work, or become too costly to keep replacing?

    1. Destroy All Monsters Silver badge

      Nanoplague to the rescue!

      Offering that Asia will fail to become pre-eminent because motherboards are rotting in their racks must be the last refuge of the Old World.

      They will have to weather serious bubble popping of the Chinese Economy though, it will be interesting.

  5. Destroy All Monsters Silver badge

    El Reg says: There's that almighty dollar again.

    No, it sounds like economic calculation. Do we want it, what will it cost, what will it bring? If the ROI isn't there or there is uncertainty about it in the first place, there will be doubts. And in the current economic climate, full of uncertainty, spastic government activity, money printing, debt craters and a "pivot to asia" that is not the one Obama is pushing, and the US and Yurop economies in the crappers of stagflation and not going anywhere soon (except for the nasty, refusing-to-be-poor-in-solidarity Germans, maybe) uncertainty is large indeed.

  6. Nigel 11


    Anyone else pondering what might be labelled 802.3BT and 802.3bu ?

  7. Pascal Monett Silver badge

    400GB/s ?

    And to think that my home network is still locked at 100MB/s because ISPs don't use GB-compatible routers...

    1. Carl Carter

      Re: 400GB/s ?

      Mine does - 1Gb down, 500mb/s up ftw! (Yes - that is wan speeds!) - not available in many places yet but it's expanding!

  8. Anonymous Coward
    Anonymous Coward


    Doesn't have the money for R&D ?

This topic is closed for new posts.

Other stories you might like