How fast are those packets moving?
blog.wesleyac.com
blog.wesleyac.com
excerpt:
' "We're having a problem sending email out of the department."
"What's the problem?" I asked.
"We can't send mail more than 500 miles," the chairman explained.
I choked on my latte. "Come again?"'
Regular electrical signals can propagate just as fast in copper wires as light does in fiber:
https://en.wikipedia.org/wiki/Wave_propagation_speed
Also, if there are any wireless (microwave) links in the path, those will actually be (almost, due to the atmosphere) at the speed of light.
- They don't factor the time the opto-electrical converstion takes at each hop.
- They don't factor queueing and buffering in any devices
- They assume the router was able to instantly generate the TTL exceeded message at each traceroute hop.
- They don't seem to consider that the return path might be going via Japan with a hop through France, which won't show on the traceroute at all.
- They assume each non-passive device on the path will show in the traceroute (this isn't true, L1/2 devices won't show, technologies like MPLS can obfuscate entire network cores).
In terms of how fast light moves in single-mode fibre it's fairly constant, the variance here ins't to do with the fibre.
Which is basically what the article is about.
The speed of packets in fiber is exactly the one calculate doing the math combining together the speed of light and the refraction coefficient of fiber.
What slowndown everything is reaching an hub, detect the light and trasmute those impulses into bits, decide where to redirect the packet, and re-trasmute the bits into light impulse.
It is not the light domain the one that dominate the speed of packets but is the electronic domain.
Am I wrong?
http://uk.businessinsider.com/plans-for-high-frequency-tradi...
I also reckon that a modern router can forward a packet in a microsecond. Once you don't involve software in the data plane, routing a packet should only take a few clock cycles.
What? How'd that be possible?
An example: http://oklo.org/2015/01/01/lightspeed/
In addition to the refraction index, you’d expect fibre optics to have another limiting factor: scatter. As light bounces off the side of the fibre, it gets scattered back and forth. I.o.w.: you’d expect a single, “perfect” pulse to eventually reach the other end as a slightly drawn out pulse (with its intensity spread out over a bell curve?). Is that true? If so, does that mean that in fibre optics , the longer the cable, the lower the frequency at which you can send pulses, to avoid “bleeding” from one into the next? And is it true radio doesn’t have that?
Or does fibre not work that way?
No, it does.
> Surely it can't go round any bend, then, though?
No, if you bend the fiber too much the light leaks out into the cladding.
Fiber optics have lots of limiting factors: power limits, chromatic and polarization mode dispersion, spectral attenuation, manufacturing imperfections, crosstalk, ...
Fiber optics are just better than a lot of other mediums, but that does not mean there aren't limitations.
- travel time on fiber
- transceiver conversion time
- time spent sitting in some buffer. This can dwarf other components on "common" networks!
http://www.m2optics.com/blog/bid/70587/Calculating-Optical-F...
Undersea fiber cables have repeaters embedded in them. So even though it's only one hop from an L2 perspective, the light is being decoded, error corrected, and re-transmitted many times along the way to hong kong from SF.
I'd say it's more likely to be some sort of routing or even filtering overhead between the two national jurisdictions. Plausibly there be some kind of layer 3 filtering going on ...
EDIT There's a good discussion of what you're talking about here [0]
Upon further reading, there's a such thing as an all optical regenerator which would appear to provide the best of both worlds, though there's nothing here about how widely these are used [1]
[0] http://www.rfwireless-world.com/Terminology/Optical-Repeater...
Yes.
> If so, then wouldn't an analog signal amp such as this simply amplify the noise along with the signal?
Yes. That's why there are limits on the reach of amplified networks.
> Decoding and re-encoding would have the effect of 'cleaning' the signal in this case.
Yes, but regenerating the signal introduces complexity, and thus reduces reliability, adds latency and only works for a specific encoding, which makes upgrades impossible. None of these things are wanted at the bottom of the sea for the next 25 years.
> If it does get noisy, where does the noise come from?
There are lots of noise sources. See: http://www.svphotonics.com/pub/pub029.pdf
> The medium?
Yes, partly.
> Leakage from outside the cable?
No. Although vibrations can cause noise.
No, all long haul subsea fiber optic cables use all optical amplifiers. They are just colloquially called repeaters.
Most of the design of optical networks traces back to 90's telco requirements and not to computer networking and during that time telcos were extremely sensitive to latency (the idea being that when the latency of the digital part of the network is smaller than some limit the analog-digital transition can be done by somewhat cheaper hardware). The funky interleaved SDH/SONET frame formats are motivated by ability to implement muldexes that introduce less than entire frame (ie. 125us) of latency.
No, this is incorrect. Subsea fiber cables use all optical amplification at each repeater.
There is no optical-electrical-optical conversion.
This is not right. Submarine cable along the ocean floor is strictly Layer 1 ,there is no L2 down there. Also nothing is being decoded and retransmitted along that path.
Signal regeneration only happens at the terminal stations. There's no repeaters along the ocean floor. The signal undergoes no conversion from optical to electrical, its optical the whole time it's under water.
Submarine cables use an optical amplifier. Basically the fiber itself is doped with erbium creating an EDFA - erbium doped fiber amplifier. See:
Isn't that what "only one hop from a L2 perspective" would mean?
But my main point was that the undersea spans are not in fact "repeatered" and there is no conversion of signal once it leaves land.
And as others also pointed out its an optical amplifier, there is no "decoded, error corrected, and re-transmitted" being done along the path of the submarine cable.
Only if you're being obtuse and don't actually work in networking. A hop in L2 is a device that does L2 forwarding and participates in L2 protocols like STP.
>And as others also pointed out its an optical amplifier, there is no "decoded, error corrected, and re-transmitted" being done along the path of the submarine cable.
This is only true for newer stuff. Look up OEO fiber repeaters. Those cables are still in use.