Sitting here on the plane on my way back from London. A number of talks regarding all the activities in the Ethernet universe, and of course the next speed. After one of these talks one individual looked at me starry eyed and said, "My head is ready to explode!" I laughed and said, "Imagine how I feel."
The simple reality is that there is a lot going on. The IEEE 802.3 100 Gb/s Backplane and Copper Cable Study Group is pushing forward with the hope of becoming an official task force in September, but before it can do that it must agree on the reach objectives for the backplane and the copper cables.
This is always a tricky one, but trickier this time than prior efforts--as people looking at NRZ are looking at approximately 25dB insertion loss at about 13GHz. Board materials will play a pivotal role as always and, with that comes the cost debate. My forehead is still stamped with "High Volume Manufacturing" from the number of times this was discussed during the original Backplane Ethernet project.
And lurking underneath the surface—the modulation debate! This time it appears more companies are considering PAM-4 than when the original Backplane project specified serial 10 Gb/s operation across a backplane.
Oh but that is not it.
The IEEE 802.3 Ethernet Bandwidth Assessment Ad hoc is ramping up. Been going all over the world discussing this, and the reality is that people are worried about their future bandwidth needs. Individuals from the financial community will be stepping up first to share their insight.
100 GbE optics development is ramping up. There is talk of a new IEEE 100GbE optics project. I expect a push to reduce the number of fibres in the multi-mode 100GbE solution from 10 to four (in each direction).
Also, what will happen with the single mode family? Advocates of an MSA-developed 10x10 2km solution may point to cost reductions with a shorter reach, but that argument ultimately didn't really fly during the IEEE 802.3ae project, which developed 10GbE. There were reach objectives for 2 and 10km, but it was determined that one solution to cover both reaches was the best decision.
So will efforts for single-mode implementations instead focus on reducing the width of the electrical interface to a 4x25 solution, and perhaps consider a non-retimed or partially retimed solution? This would remove the inherent 10:4 mux currently needed. This would drive cost and power down while driving density up. Maybe, but can a non-retimed solution actually be developed that supports the current 100GBASE-LR4 specifications, or will a shorter reach be necessary in order to enable a non-retimed solution?
Ok, is that it?
No, not by a long shot. Hearing rumblings about next-generation signaling over twisted pair. That should be a fun debate. Oh, also hearing about POE++--yup "MORE POWER." I keep thinking the POE community should consider get Tim "The Toolman" Taylor to promote it.
With all this said, I am looking forward to the Ethernet Alliance’s Technology Exploration Forum. The agenda includes panels addressing many of these topics, and it will be interesting to see where the discussion leads. For more information go to the forum's website.
Oh, and I shouldn’t forget to mention the IEEE 802.3 Interim meeting coming up in May.