The data capacity of a fiber optic cable

September 9, 2013 6:54 PM

How much data can you shove through a fiber optic cable? Quite a lot, as it turns out. Using the ITU standard 50 GHz dense wavelength division multiplexing grid, not only can you easily do 80 x 10 Gbps channels in a single fiber pair, but recent advances in modulation technology mean that with QPSK, 4QAM or 16QAM modulation, 1/80th of a dark fiber pair can carry a 100 Gbps signal in the optical space previously occupied by a single long distance 10 Gbps circuit.
posted by thewalrus (39 comments total)

And yet this YouTube cat video is still buffering dammit
posted by Doleful Creature at 6:57 PM on September 9, 2013 [24 favorites]

Speaking as an Australian, I will literally never ever know.
posted by turbid dahlia at 7:03 PM on September 9, 2013 [18 favorites]

I'm poor AND in the middle Midwest, so I am with turbid.
posted by Samizdata at 7:07 PM on September 9, 2013 [1 favorite]

There's the guy who set up a 40 gigabit ethernet link to his 75 year old mother's house as a proof of concept of what you can do with residential dark fiber. Only temporarily of course, in 2007 a router capable of handling a 40GbE line card was probably as valuable as the house itself.
posted by thewalrus at 7:12 PM on September 9, 2013 [1 favorite]

I was already pretty mind-boggled when I saw a cross-section of an undersea fiber-optic cable and realized how tiny the actual data-carrying portion of it is.
posted by mstokes650 at 7:13 PM on September 9, 2013 [13 favorites]

That's a near shore, heavily armored cable, it gets thinner in the deep ocean segments... But yes, if you pull an individual tube out of a loose tube singlemode fiber cable, that tube will contain 6, 8 or 12 strands which are miniscule in diameter and almost impossible to keep track of visually. The individual strands in the tubes are 250um in diameter when cladded, so 1/4th of a millimeter.
posted by thewalrus at 7:15 PM on September 9, 2013 [4 favorites]

The Corning glass museum's display of how much copper it takes to match a single fiber is a great visual for this - and obviously pre-dates these technology improvements!
posted by meinvt at 7:21 PM on September 9, 2013

How does it compare to a station wagon full of magnetic tapes?
posted by Harvey Kilobit at 7:35 PM on September 9, 2013 [9 favorites]

Just in case you didn't read it at the time, probably the best thing ever in Wired magazine, Neal Stephenson's "Mother Earth Mother Board", about the laying of the FLAG project's undersea cable network.
posted by sutt at 7:47 PM on September 9, 2013 [5 favorites]

How does it compare to a station wagon full of magnetic tapes?

...or micro SD cards?
posted by pompomtom at 7:56 PM on September 9, 2013 [5 favorites]

I was impatiently waiting for a disc image to download because my connection was a little slow. Over 8 minutes! Lying ass dogs at Comcast! Then I remembered how long it would have taken just a few years ago, much less over dial-up.

I know that this isn't about residential speeds, but if it was, I wonder if I would ever get that impatient again. I can't think of anything that I could ever want to download that would take more than one second at that speed.
posted by double block and bleed at 7:57 PM on September 9, 2013 [2 favorites]

I have no idea why I've watched two of the videos that thewalrus linked, but I now know how to strip loose tube and armored fiber optic cable. I'm about to learn how to terminate the fibers. I expect this knowledge will never, ever be relevant to my life in any way, whatsoever.
posted by Ivan Fyodorovich at 8:05 PM on September 9, 2013 [4 favorites]

I was already pretty mind-boggled when I saw a cross-section of an undersea fiber-optic cable and realized how tiny the actual data-carrying portion of it is.

It is the small diameter of the optic fiber that makes it work by restricting transmission to a single-mode/single ray of light. This produces optimal bandwidth and transmission distance characteristics. The glass fiber is about the diameter of a human hair and the data carrying core of the glass fiber is less than one-tenth of that.
posted by JackFlash at 8:06 PM on September 9, 2013 [1 favorite]

Of course the downside is that advances in network capacity have the potential to be misreported as increases in network traffic. Those reports could eventually go viral and become blindly accepted by the mainstream press as common wisdom about the growth of Internet use. Finally the published articles could become justification for thousands of startup business plans and VC pitches. Next thing you know, the DOW is over 15 thousand and companies like Yahoo! are in the news again.
posted by ceribus peribus at 8:07 PM on September 9, 2013 [4 favorites]

As I understand it, the nice part about fiber is that it just grows in value once laid -- as new multiplexing comes along, the cables themselves don't generally need to be upgraded, just the end points pumping light into them and reading it back out on the other end.
posted by pwnguin at 8:09 PM on September 9, 2013 [4 favorites]

pwnguin, that's generally correct. We are still using first-generation singlemode fiber that was laid between Boston-New York in 1985-1986. They have replaced the optical terminal equipment and routers on each end of the link many times since then, growing in capacity from 155 Mbps per circuit all the way up to 40 and 100 gigabit Ethernet.

Fiber, once in the ground, has a service life greatly exceeding 25 years as long as it does not meet with any backhoes... Even if it does, the damaged section can be removed and new cable spliced in place in less than 24 hours.
posted by thewalrus at 8:19 PM on September 9, 2013 [2 favorites]

How does it compare to a station wagon full of magnetic tapes? pretty well, actually, if you substitute "magnetic tapes" for "4TB hard drives"... The time that it takes to write several hundred terabytes of data onto empty 4TB drives is significant, as each drive can only sustain a write rate of approximately 75 to 105MB per second, then the transport time in the station wagon, and then the time it takes to read 4TB of data off each drive (even if done in parallel), at 100-115MB/second best case scenario.

100 or 200 Gbps end-to-end over a network such as between two research facilities with huge, fast SANs would be faster.
posted by thewalrus at 8:21 PM on September 9, 2013 [1 favorite]

So 1/80th of a fiber can carry 100Gb/sec, for a total of 8,000 Gb/sec.

Actually the state of the art is about double this.

1/44 of a fiber can carry 400Gb/sec for a total of 17,600 Gb/sec.
posted by eye of newt at 8:26 PM on September 9, 2013

My favorite trick is when you treat a giant fiber loop as a hard drive. They did that in Canada -- bunch of nodes just passing data around, bucket brigade style, changing whatever they needed to change as they passed the signal around. They quite literally had a nationwide hard drive whose capacity was established by the speed of light.

I believe it was hundreds of GB back in the day; it'd probably be tens of terabytes now. Not much, until you realize it existed countrywide.

(Yes, there are better ways to do this. Not sure there are cooler ways, though.)
posted by effugas at 8:28 PM on September 9, 2013 [9 favorites]

It's not really specified in that AlcaLu link what the bandwidth of each wavelength is within the dark fiber, though it says 44 wavelengths... I am assuming that it's 16QAM modulated "super channels" which are 100 or 200 GHz wide and cannot operate in parallel with existing 50 GHz DWDM on the same dark fiber. Assuming also (it doesn't say) that the 44 wavelengths are all in the optimal lowest-attenuation band around 1550nm frequency.

In 50GHz wide channels, 1/80th of a fiber can carry 100Gbps using Cisco's QPSK modulated linecards, or can carry 200 Gbps with 16QAM modulation. Therefore 200 Gbps * 80 = 16000 Gbps.

The DWDM terminals and router equipment on each end to actually achieve 10000 Gbps+ would cost as much as the GDP of a small nation.
posted by thewalrus at 8:29 PM on September 9, 2013

Do you have to replace the repeaters to use these new encoding techniques? In other words, do you have to lay new fiber?
posted by Chocolate Pickle at 8:37 PM on September 9, 2013

EDFAs (erbium doped fiber amplifiers) on a path generally do not need to be replaced, nor does the fiber need to be replaced. If all you doing is flat amplification of the signal such as across the band from 1520 to 1610nm every 80 km, no new electronics between the end points. If for some reason you have equipment doing 3R (reshape, regenerate, retiming) the retiming part would require replacement.
posted by thewalrus at 8:39 PM on September 9, 2013

Me: [reads post out loud] It's about bandwidth! You should comment!
Network engineer husband: You have no idea how much of that I understood... but it's pronounced "quom." I'm glad they included 100Ghz, though; that's the latest and greatest.
Me: Are you being facetious?
posted by Madamina at 9:15 PM on September 9, 2013 [4 favorites]

I was already pretty mind-boggled when I saw a cross-section of an undersea fiber-optic cable and realized how tiny the actual data-carrying portion of it is.

I'm still not quite able to wrap my mind around the fact that there can be transoceanic cables, and we've been doing that since 1858. "Oh, no big deal, we just tied two continents together across an ocean with a piece of string." What the hell, humanity? You're kind of awesome.
posted by jason_steakums at 9:32 PM on September 9, 2013 [8 favorites]

For SAN replication I'm looking forward to a future where we've solved the problem of distance delay...
posted by Annika Cicada at 10:14 PM on September 9, 2013

jason_steakums: "I'm still not quite able to wrap my mind around the fact that there can be transoceanic cables, and we've been doing that since 1858. "Oh, no big deal, we just tied two continents together across an ocean with a piece of string." What the hell, humanity? You're kind of awesome."

When you think about how much telegraph wire connected the US cities in 1860, the real challenge seems to be making them seaworthy. And maybe that you can still read the signal on the other end without needing a nuclear power station to power through it.
posted by pwnguin at 11:25 PM on September 9, 2013

When you think about how much telegraph wire connected the US cities in 1860, the real challenge seems to be making them seaworthy. And maybe that you can still read the signal on the other end without needing a nuclear power station to power through it.

Or blowing through the insulation, which is what happened to the first cable laid across the Atlantic (instantly destroying a huge investment - they didn't have the technology to fix cables yet).
posted by atrazine at 1:36 AM on September 10, 2013

Nice. So now that's there ton of dark fiber (fiber laid, but not used) in the US and probably Europe, at least in major urban centers. Where is my 1 gig fiber to cabinet for $5 a month?
posted by elpapacito at 5:33 AM on September 10, 2013 [1 favorite]

I was impatiently waiting for a disc image to download because my connection was a little slow. Over 8 minutes! Lying ass dogs at Comcast! Then I remembered how long it would have taken just a few years ago, much less over dial-up.

In my experience, these issues aren't the fault of the ISPs, but of the other end. I remember back in the early days of my cable internet connection, I was generally underwhelmed by the speed of downloads. Then I had to download something from Compaq's (I'm old!) ftp (I'm old!) site. Unbelievably, blazingly fast. I actually saw the full rated speed of my connection. And further tests confirmed this. Nearly all sites would max out at some speed.

There was a thing on here (or reddit?) about youtube content caching. That if you disabled the netblock that the cache was located on, your browser would have to get the content right from youtube.com, and that made it tremendously faster.

Nice. So now that's there ton of dark fiber (fiber laid, but not used) in the US and probably Europe, at least in major urban centers. Where is my 1 gig fiber to cabinet for $5 a month?

Sadly, that fiber isn't where it needs to be for what we want. My understanding of the dark fiber issue is that when they take the time to survey and trench for a fiber run, they include lots of extra pairs of actual fiber, because that costs almost nothing compared to the labor costs of putting it in the ground. So the dark fiber only goes places where fiber already goes.
posted by gjc at 7:27 AM on September 10, 2013

My favorite trick is when you treat a giant fiber loop as a hard drive. They did that in Canada -- bunch of nodes just passing data around, bucket brigade style, changing whatever they needed to change as they passed the signal around. They quite literally had a nationwide hard drive whose capacity was established by the speed of light.

I believe it was hundreds of GB back in the day; it'd probably be tens of terabytes now. Not much, until you realize it existed countrywide.

(Yes, there are better ways to do this. Not sure there are cooler ways, though.)

Was there actually more data than could exist in the buffers of the switching equipment?

Another trick with fiber was when someone figured out that a coil of fiber could be used to make a very accurate gyroscope. With none of the delicate moving parts. The light going through the coil always takes the same amount of time to go through the 360 degree coil, so position can be determined by measuring any time/phase difference between the known location of the signal source and the unknown location of the receiver. (Metaphor: if you put a pebble onto a merry go round, and someone else picks it up later, they can determine your position if they know when you placed it. Or they can determine when you placed it if you know their location. Because the speed of the merry go round is constant at the speed of light in the medium.)
posted by gjc at 7:45 AM on September 10, 2013

> My favorite trick is when you treat a giant fiber loop as a hard drive. They did that in Canada -- bunch of nodes just passing data around, bucket brigade style, changing whatever they needed to change as they passed the signal around.

[...]

(Yes, there are better ways to do this. Not sure there are cooler ways, though.)

Have you heard of PingFS?
posted by The Lurkers Support Me in Email at 9:13 AM on September 10, 2013 [1 favorite]

When you think about how much telegraph wire connected the US cities in 1860, the real challenge seems to be making them seaworthy.

Field: Do not listen to this man! Should the cable descend below the surface of the sea, it shall surely be devoured by the many fishes!
Smuckles: A problem easily solved. We will wipe the cable with dogs, so it becomes offensive to all creatures.
Field: Hm. Mr. Smuckles' reasoning has given me pause. Yes, let us use dogs.
Brunel: I have built a steam ship which can house ten thousand mongrels! Your endeavor shall not want for the horrors of the canine body.
posted by FatherDagon at 10:23 AM on September 10, 2013 [1 favorite]

effugas: " They quite literally had a nationwide hard drive whose capacity was established by the speed of light."

To be fully accurate: the speed of light in glass, reduced by the cosine of the acceptance angle of the fiber, more or less (different multiplexing schemes complicate the numerical aperture issue) - in terms of the physical distance travelled, or simply c/n for the distance along the bouncing light path.

But, yes, something close to c.

/pedantic optical engineering mode off.
posted by IAmBroom at 10:58 AM on September 10, 2013

gjc: "In my experience, these issues aren't the fault of the ISPs, but of the other end."

This is usually true, but it had been slow all day.

There was a thing on here (or reddit?) about youtube content caching. That if you disabled the netblock that the cache was located on, your browser would have to get the content right from youtube.com, and that made it tremendously faster.

I would love to have a link to this.
posted by double block and bleed at 11:20 AM on September 10, 2013

I would love to have a link to this.

This, probably.
posted by inigo2 at 11:51 AM on September 10, 2013 [5 favorites]

Came here to say cool things about fiber optics, saw I was beaten to it (hello, PC-1!). Leaving with great info about YouTube buffering. Thanks, MeFi!
posted by pointless_incessant_barking at 12:53 PM on September 10, 2013

It's astounding that we can figure out ways to move data ever speedier, but we can't figure out how to end-around the greedy TWC's and Comcast's of the US.
posted by yoga at 2:06 PM on September 10, 2013

sutt: "Neal Stephenson's "Mother Earth Mother Board", about the laying of the FLAG project's undersea cable network."

That article was very long but very well worth it.
posted by double block and bleed at 4:27 PM on September 10, 2013

yoga: "It's astounding that we can figure out ways to move data ever speedier, but we can't figure out how to end-around the greedy TWC's and Comcast's of the US."

We know darn well how to do it, we just won't, because socialism or something.
posted by wierdo at 4:44 PM on September 10, 2013