2:17 PM -- While doing some conference planning, I noticed that Ethernet inventor Bob Metcalfe is going to be talking at OFC/NFOEC on the subject of Terabit Ethernet.
Yeah. Terabit. With a "T".
His speaker abstract gives a pretty effective tease: "I want to talk not just about how we got to 10-Gigabit Ethernet (10GbE), but also about how we are going to get to 40-, 100-, and, yes, even 1,000-Gbit/s Ethernet, which I hereby call Terabit Ethernet (TbE)."
So, I have questions for Mr. Metcalfe (and I hope to spring these on him in person):
Warming up: Do you feel there's a market forming for Terabit Ethernet (TbE)?
What will make that market? How will we know it's time to build those products?
Who's going to buy TbE? Seriously.
What's the traffic driver? (Maybe that's obvious, but is video really enough to make this a market in the next decade?)
What kind of infrastructure is going to be needed to support TbE? We're running 100-Gbit/s Ethernet over the same stuff that 10-Gbit/s Ethernet was built on, in many cases. How is that going to change?
How will the Ethernet standard itself have to change to adjust for the packet sizes required for Terabit Ethernet?
What's going to happen to Sonet if Terabit Ethernet really is around the corner?
That's a start. If I get my shot at Bob, what would YOU like to know?
re: Terabit Ethernet But then as a minimum, it could infuse $100Ms of R&D investment for some really cool technology work. At least now, aggregate access bandwidth is pretty high, unlike the Terabit hype in the '90's when everyone was still on dial up... Still, I'd like to know where he sees the BW inflection point that will drive this.
re: Terabit Ethernet Carriers are headed to Ethernet because the Enterprise is basically fully Ethernet. If there is no Enterprise Market for Terabit Ethernet, there is not likely to be the assured market for chip makers to get devices in place for it.
With the buy in of the Enterprise, there is a market to drive down costs with much higher margins than one gets in the carrier business.
I doubt they would go near it. Take away the message that it is a technology that everyone knows and understands and the only real value left is the packet format and lack of need to translate between transport infrastructure and systems.
I would be interested to know to what extent the packetization is what Ethernet has become? We see Ethernet packets carried in MPLS across Ethernet transport using MPLS as a control plane. If we can drive the capacity then that's great, I can imagine that we might need to think again about how long the systems (routers) connecting these links can handle the volume of control plane traffic being carried. I am not a fan of PBT but it points the way towards the control plane needing to become decoupled from the forwarding plane, meaning that Ethernet just becomes a packet encoded on to a wire at higher and higher capacities.
Offline path computation and signalling should allow for data connections to be more temporary and allowing Resource Admission Control and other IMS related SOA architectures. The world is a million miles away from truely delivering all of this and the longer router vendors own the control plane the longer it will take. Perhaps Terabit Ethernet will pave the way for generic hardware and introduce new entrants into the market for control plane software.
Machine learning is primed to help service providers run more efficient and effective networks, but first the good ideas have to make their way from the lab to the real world – and that's a big challenge, according to the University of Chicago's Nick Feamster.