400GbE - Ethernet Alliance Study Group

As mentioned on the Website of the Ethernet Alliance, a new study group is established which will take concern of 400 Gigabit Ethernet.

It’s quite funny… 40GbE is just at finding its way into datacenters and carriers which aren’t top-notch and blessed with budgets, 100GbE is really hard to find, and the experts are already talking about the successor. Of course, equipment working and scaling to and around TBit, it’ll just be a matter of time when the new gear hits the market. At prices in the orbit, imho. Some of the questions which arise togehter with that announcement are:

• What’s the maximum distance over whicht the transport will be working
• Will already existing cabling be usable with the new standard
• Are the connectory suitable for this speed boost
• What about the TDP

New cabling will be – hopefully – already OM4, taking the challenge with

300m at 10GbE, but what about 400GbE? Will the dispersion of LC-connectors be low enough to realize that speed? And the last one, how’s the TDP. I suppose that the resulting themal energy is quite high, so SFP+ won’t be the factor of choice. Will there be a revival of XENPAK? The surface of XENPAKs is unequally higher than the SFP+-pendants, but again, it’s just a matter of time and technical evolution. Smaller structures in the semiconductur production process and higher integration will bring us those high port concentrations some day. In times of NVGRE, VXLAN and comparable cloud networking infrastructure where the paradigma of “resources usable everywhere” presents us a loose control about the traffic flow, SDNs and tight interaction between networking hardware and at least some kind of intelligence the stop or go will be the available hardware and transport technology.

It’ll be a thrill how technologies that build upon a lower transport will scale. MPLS/VPLS will possibly gain more importance, as it’s the logical consequence of bandwidth and port re-use. Giving away one port for just one purpose… well, good bye collapsed backbone, hello concentrated core again. More intelligence at the edge and at customer facing ports, and lower knowledge but higher speed in the levels upwards. Thinkably the price per port won’t be comparable to nowadays prices for 10GE. How long did it take for 10GE to find its place in small and midsize data centers or carriers? Compared to 1GE the time was about doubled. But nevertheless a really exiting topic, I’ve already put up a follow at twitter.