Meanwhile in aus...
...NBN CVC capacity is $80000/m for 10Gb/s, down from $150000/m last year.
International bandwidth prices are plummeting, with some routes offering 10 Gbps connections for less than US$5,000 per month. That's the cost of a link between New York and London, according to international telco analyst firm Telegeography (best known for its submarine cable maps). That route, however, has always been one of …
Consumer bandwidth (GB/month) demands skyrocket, meaning that not all of those lovely outcomes will unfold quite as ideally as you may hope.
"Hey boss. The cost of our wavelength is only a third of what it was."
"Great. Call them up and get five more. This damn Netflix 4K is killing us."
But from this it becomes apparent where the growth in consumer bandwidth demand will end. As people cut the cord and cable/satellite mostly converts to streaming over the next decade, and more people start watching in 4K, eventually it maxes out. There's no benefit to going higher than 4K unless you have an absolutely gigantic TV - probably a small enough niche it wouldn't be worth the cost of delivering 8K streams (and even if they do it is only roughly a doubling of HEVC compressed bandwidth)
We're only a few more doublings away from seeing the giant yearly jumps in bandwidth demand slow to single digit percentages and eventually leveling off. At least unless/until something denser than video streaming becomes possible, like some sort of immersive VR that requires an order of magnitude more bandwidth. Which if/when it comes will mostly be used for porn, no doubt.
Humans have limited bandwidth for sensory input, and visual input is as dense as it gets.
When have I heard that argument before?
Oh yeah, the entire history of computing. 640k should be enough for everyone. We're at the physical limit for Moore's law.
Analogue TV was fine; why would you want to go digital? SDTV is fine, why would I need HD? HDTV is fine, why would I want 4k?
The web was originally text-based. Then came pictures. Then blocky video.
Basically, people will always find a way to waste extra bandwidth.
If nothing else, Windows 10 will probably require daily multi-TB updates in a few years time.
Obviously my point went woosh over your head and you resorted back to the tired "640K" argument and want to claim that bandwidth demand will increase forever. We went from text to pictures to audio to crappy video to HD video and now 4K video because we want richer sensory input. But we've reached the limit, or are near it if 8K happens.
It is clear we aren't going to go 8K, 16K, 32K and on forever because regardless of the size of your TV once the pixels are smaller than your eye's ability to distinguish pixels further upgrades are irrelevant. There was quite a race for higher resolutions on phones for a few years but that's pretty much ceased now despite the marketing points some might think there could be from selling a "4K" phone. Ever stop to think why?
Where's the drive for more bandwidth when we've maxed out what we can input to our senses? Windows updates are already lost in the noise versus streaming, so that's not it. Only if we find some sort of richer input to our senses...though if we directly wire into our brain the bandwidth of our optic nerve is less than 10 Mbit ethernet!
Consumer bandwidth (GB/month) demands skyrocket,
Most of it is to the nearest CDN Cache Node. What we see is a natural result of the CDN success as a delivery mechanism. We have reached the point where everyone is using it which rather unsurprisingly increases demand for datacenter, datacenter connectivity and broadband bandwidth.
At the same time the demand for long haul drops. By the way - moving to 4 or god forbid 8K will not change this, in fact it will make the tendency even more pronounced because delivering 8K in a usable form without a nearby CDN is unrealistic.
I guess they are both technically correct, which is the best kind of correct.
"Wavelength-division multiplexing (WDM) can transmit multiple optical carrier signals on a fiber pair by using different wavelengths. These wavelengths are sent down branches to their expected destination using wavelength filters. WDM provides full mesh connection by wavelength and facilitates the allocation of dedicated wavelengths between two specific nodes. WDM can also be upgraded by adding more optical channels without additional fiber. The multiplexing of several wavelengths on the same fiber allows 16 x 2.5 Gbps of information to be carried throughout the oceans."
So, 4 multiplexes on a wavelength gives you the 10Gbps.
They can get a lot more than 16 wavelengths on a fiber, or 2.5 Gbps per wavelength now. That's really why it is decreasing - it isn't because someone is out there laying more and more cables every day, it is because we're able to increase the carrying capacity of existing cables every few years.
Yeahbut ... wavelength is measured in meters. Bandwidth is measured in bits per second. You don't sell fiber links based on the colo(u)r of the laser, you sell them based on how many bits you can cram down the glass in a given time frame.
Or, rather, that's how it was last time I provisioned an international fiber link. Wouldn't surprise me if marketing managed to completely fuck up the meaning of yet another technical term through sheer incompetence, thus making life difficult for people with clues.
Unless the numerologists have managed to sell a certain population of the planet on "auspicious" frequencies for their data ... Wouldn't surprise me.
Jake was perplexed by "Wavelength? Not bandwidth?"
In this context, a "wavelength" can be interpreted as a "colour" (although some may be invisible IR). They're stuffing more than one wavelength down the fibre.
[I expect that there's a great deal of headroom (room for expansion) in this area. I doubt that they're efficiently filling the optical spectrum. <- Gut-Instinct Alert.]
The word bandwidth is the one that has been corrupted. It used to mean, ah, bandwidth. As measured in kHz or MHz. It was fair to apply it to bps. But now it's been corrupted to mean GB/month.
The cost of "a wavelength" is the cost of a DWDM light path on a fibre, at one particular wavelength/frequency/colour. Typically it would give you a 10Gbps or 40Gbps data path. However the same fibre can simultaneously support different wavelengths for different customers.
This is as opposed to "dark fibre", where the whole fibre is dedicated to you - you can light it with as many different wavelengths as you like, simultaneously.
AC suggested, "This is as opposed to 'dark fibre', where the whole fibre is dedicated to you - you can light it with as many different wavelengths as you like, simultaneously."
Not if the fibre is sufficiently long that it has repeaters installed (perhaps 100km). Those repeaters, being active, would be designed to work on certain wavelength(s).
Of course, on land-based installations you could wander over to the huts and upgrade the repeaters.
Weell, it is true that the erbium-doped fibre amplifiers used in modern-day repeaters restrict the bandwidth (in its original sense), but it's rather wide, typically covering 1530-1565 nm for a C-band system - that's enough to shift 10 Tb/s (10 x 100 Gb/s channels) or more across oceanic distances with modern modulation techniques.
Some fascinating research from NTT Data about the developing SDM
The exponential growth from Tbps to Pbps doesn't actually seem that far off.
Biting the hand that feeds IT © 1998–2019