I visited Fiji a number of years back. I was totally expecting the internet to be of pretty poor quality in the middle of the South Pacific - but it was fantastic - on almost every island I visited. I later learned what this maps shows... its a stopover for pretty big fiber lines from Hawaii to Australia....
They’re a composite cable, with fiber for the data portion and copper for the power needed to drive the fiber optic repeaters.
I was able to tour a fiber laying/maintenance ship back in the 1990s. Had it explained that they lay the cable powered and under end-to-end test (shore to ship). This makes the spool of cable into a giant electromagnet, meaning there were signs all over portions of the ship banning steel tools, watches, etc.
Does anyone know if the paths shown for the transatlantic cables are anything like accurate? I would have expected them to follow "great circle" routes since they are the shortest. For example, the shortest path from Long Island to the Bristol Channel goes through Newfoundland and the southern tip of Ireland. There are several cables on the map with those end points but none of them are shown as taking that route. I guess this is poetic license that ignores the distortion of the map projection. Is there a map that shows the routes more accurately?
The cable landing stations provide DC power for the repeaters from both directions. The power requirements are substantial and use thousands of volts.
All the components of the subsea cable system are engineered with reliability in mind. None of the components are to require maintenance over the lifetime of the system. This means components are expected to operate 20 to 25 years at a minimum.
> The power requirements are substantial and use thousands of volts.
While 100% correct statement, the wording here could potentially lead someone to make an invalid assumption.
The core reason why it’s “thousands of volts” isn’t because a lot of power is needed specifically (although that is the case), it’s because of the nature of electrical conductors. Power loss over long distances is much more heavily influenced by current (current is squared in fact in the equation) rather than voltage. To say another way, you’ll have a lot less power loss with 50k volts at 100 amps than you would with 250V at 2,000A, despite both having comparable wattage (this is in part why high voltage power lines exist on land). Another reason is that electrical conductors (especially with copper, but also with aluminum and others) are heavy and the higher the amperage, the larger the conductor needs to be and thus the more weight.
The extra distance for NY/NJ-Bristol cables to cover is "just" ~400-500km (extra ~7%), plus if you look at ocean floor map [1] it seems they prefer not to put the cable in the relatively shallow waters of Newfoundland or Nova Scotia (maybe this decreases possibility of a damage by anchors, or they compensate a bit in length by going deeper, i.e. shorter circumference due to smaller radius).
The other possibility is that the direct path passes underwater mountainous region with quite variable depths in mid-Atlantic [1], so going up-down or follow the valleys anyway would increase the cable length. The right answer is likely a combination of all these to minimize cable cost, maintenance and latency.
Ships anchors and trawlers are major risks for subsea cables. Deeper is thus better.
Underwater geography is also a major consideration. It's not just the going up and down, it's the risk of landslides, tectonic shifts, volcanoes, currents and icebergs scraping the ocean floor.
On top of that are geopolitics and enviromental concerns.
1901 telegraph submarine cables [1] were more straight for a similar route. Likely because it was essential to make them as short as possible going underwater, so most of them were going from Ireland to Newfoundland.
Facebook operates a vast global network connecting its datacenters and points of presence (POPs). When you connect to Facebook from, say, Tokyo your HTTPS connection is terminated in a local point of presence (likely in Tokyo or somewhere in Japan) and from there the connection is forwarded on to a datacenter in the US or Europe. This forwarding happens over Facebook's privately operated network backbone, which needs to be substantial enough to handle billions of users' traffic (not to mention internal replication traffic and so on). Renting capacity from commercial providers is expensive so Facebook and the other big tech companies are interested in building their own infrastructure for this purpose.
Of course, it would be possible to have a user in Tokyo connect to Facebook's network somewhere closer to the DC but a) getting onto FB's high quality, low congestion network ASAP will perform yield much better performance than leaving the connection running over the internet where it's subject to congestion and long SSL connection round trips, b) terminating user sessions in POPs allows much finer control over which datacenter users are ultimately sent to. Without this FB would have to rely on DNS changes to steer traffic and they are slow to propagate and crude.
Facebook is responsible for about 8% of all Internet traffic. They need subsea cables to connect their datacenters and to deliver traffic to end users.
A submarine cable can't just connect two end points by the great circle route. A cable route has to consider subsea geographics and geopolitics. This includes volcanoes and congress. Either one can be a big obstacle.
What's the location with the greatest connectivity? It's not New York City. Not Hong Kong (though it could be number 1 if you include some very near cities). And amazingly, San Francisco and the Bay Area have no cable landings at all near them.
It looks like it's Fortaleza (Brazil) with 17 cables making landfall there.
Interconnection facilities are the best places to be (or rather what you want to be connected to in order to establish as many adjacencies). 111 8th in NYC. 165 Halsey in Newark. 350 easy cermak in chicago. Palo alto for the bay (but also eqx in San Jose). Infomart in dfw. Etc.
Cable landing stations generally service one cable and the actual networks that use it are generally some distance away.
I thought it was Frankfurt in Germany, but can't confirm, there is only public data on per country connectivity, where Germany is at the top. But Brazil is definitely in top 5, for top 5 connectivity from last year the countries were: Germany, Brazil, Ukraine, U.K., Switzerland. Submarine cables are only a small part of connectivity.
Correct, Frankfurt and Amsterdam are densest submarine fiber endpoints in Europe at least. FRA is EU's main financial and business hub, and even AMS, which has huge bandwidth, serves only as a transit location for FRA fiber.
Frankfurt is about to become Europe's datacentre alley. Dublin is ahead now, but they will take over pretty soon
Biggest bandwidth is between North America, Virginia more precisely and Europe.
There's datacentre alley in NOVA, probably tapped by three letter agencies (why would they, Loudoun and neighbouring counties, have most internet capacity in the world then?). About 70% of total internet ttaffic or more goes through NOVA
That 70% doesn’t pass the sniff test given half the world loves on the opposite side from there. No communications between Europe, Africa, Asia Or Austrailia would route via the east coast, neither would most traffic from North America to east Asia, not to mention all the intra-regional traffic.
Singapore is in the middle of it. It seems more fair to look at a general area instead of an exactly coinciding point on the map (it's not going to physically coincide on the same point anyway).
The general NJ-NYC area seems to be pretty dense with cable landings.
Lots of cables going through Singapore & Johor (Malaysia) - really hard to make out any details. I've dealt with abandoned cables in the area and the map doesn't look very accurate. Would have liked to be able to zoom in more.
Old but gold. I love this map because it reminds me to pack my stuff and move to Tulum or Cancun to start operating my company from there as they have submarine cables landing there
though this doesn't show military cables, it reminds me a good book [1] about US eavesdrop operations against Soviet Union by literally surrounding their underwater cables with a special kind of "ring" that tapped electronic signals under it. (Fun fact was that USSR did not even encrypted those communications)
[1]: Blind Man's Bluff: The Untold Story of American Submarine Espionage by Sherry Sontag et al.
Thanks for that, I read that book when it came out but forgot the name. Every time the cable map was linked I think about that book. Some great stories in that book about tapping the cables and how used to be a relatively common occurrence for subs to crash into each othe because they were running silent.
Dark docs is kind of watered down history. The previous mentioned book is great. Back in the 90s Discovery actually had documentaries that covered this from the people actually involved (including the person who proposed the operation) - now we have SHARK WEEK.
Some very offbeat places like Papua New Guinea, Fiji, and Madagascar(!) have lots of cable landings. Would it be safe to say that they have excellent Internet connectivity? I mean good bandwidth and reliable service, although latency to Europe or the USA may be high due to distance.
Look at what Saint Helena is going through. They're kicking in a bunch of extra bucks so that their island is a landing point for a cable under construction, but the minimum spend is for a 200gbi/s connection.
The preponderance of interconnects is a necessary but insufficient condition for good internet. A good number of those lines will be allocated as dedicated circuit, private dedicated lines, and other purposes.
Speculation, but I would guess no. Having good service means you need an ISP that provides last mile service. You can't just directly plug into the undersea cables, and I doubt the places you listed have good ISPs
Also understated is that they are essentially transit points, so the bulk of bandwidth is dedicated to bigger places (more Asia-America and less in between).
This is fascinating and leaves me with so many questions. What do these cables look like? How are they made? How do they lay them? Are they in sections? How are they connected? How much maintenance do they require? How much data can they carry?
I am so glad you asked! Neal Stephenson answers all your questions and then some in the timeless "Mother Earth Mother Board" https://www.wired.com/1996/12/ffglass/
Fiber theoretically has unlimited bandwidth. The bandwidth of these cable system is limited by the transmission gear used to light the fiber. This is the same for all optical telecom gear though. The cables themselves have a lifespan though. Newer systems have a lifespan of around 25 years.
Cable systems have a design lifespan. This is usually around 25 years.
The cables themselves have a lifespan, but it's far longer than 25 years. Many cable systems have had their lifespan extended beyond 25 years, but is more of a question whether they are economically viable, rather than technically viable.
A lot of cable systems have been retired, purely because they had become obsolete and a new system was more cost effective.
Fiber optics cables themselves are very longlived, if there is no external trauma. There are cables from the 1980s still in use. It's more likely that the amps in a subsea system fail long before the cable. Once the amps start failing, it usually isn't cost effective to keep using the cable system anymore.
>"Cable systems have a design lifespan. This is usually around 25 years."
Literally the last sentence of my post is "Newer systems have a lifespan of around 25 years." What is your point?
>"The cables themselves have a lifespan, but it's far longer than 25 years."
No this is incorrect. There are a many factors that limit the lifetime of "wet plant" - chief among them are fiber hydrogen exposure, which causes attenuation. There is also glue and insulation aging as well as corrosion. And "the cable" in submarines systems always includes all of the passive optical components, alignments, joints and passive electronic components. Extending these system is generally on the order of 5 years. That could hardly be classified as "far longer than 25 years."
I really enjoyed an episode of Mighty Ships (episode 1:7 -just looked it up) showing just how hard laying cables is. A surprise to me is they carry quite high voltage to power seabed repeaters. It shows them at one point trying to find a short in the massive cable.
I don't have a link handy for you but search for "Hibernia Atlantic" on Youtube. They've posted many videos over the years showing exactly how it's done.
Where do they connect to from the sea? How do they secure those locations? Are they all privately owned? How much “public” vs “private” traffic is on them? So many questions
They go very deep underground (rather than lying on the seafloor) quite a distance out from the coastline, and stay deep underground until they get to the "landing manhole" which is nowhere near the beach. The manhole is generally surrounded by a security perimeter. The "landing station" where the NSA's equipment sits is either in the same compound or (less ideally) very nearby. In the latter case the cable run from the manhole to the landing station has to be protected as well.
The digging that they do on the beach is quite disruptive. It often involves blasting. Beachfront landowners fight these things tooth and nail, which is why the landings tend to be in either highly industrialized port locations (Portland, OR) or middle-of-nowhere rugged rocky beaches that nobody visits or lives near.
>"They go very deep underground (rather than lying on the seafloor) quite a distance out from the coastline, and stay deep underground until they get to the "landing manhole" which is nowhere near the beach. The manhole is generally surrounded by a security perimeter. The "landing station" where the NSA's equipment sits is either in the same compound or (less ideally) very nearby. In the latter case the cable run from the manhole to the landing station has to be protected as well."
This is not true. Cable is only buried when the depth is less than a 1,000 meters. At 1,000 meters and above it just sits on the sea floor.[1] There is also no "manhole" where a cable pop's out of the ocean floor. The NSA does not need to be in a cable landing station, there is no benefit. The wavelengths can be peeled off fibre optic cable anywhere along the path. Hence why they were doing this at an AT&T Central Office on Folsom Street in S.F.[2], nowhere near a cable landing station. Cable landing stations are actually pretty uneventful places, generally some nondescript industrial park. The security there is the same as it is at any IX or data center.
Some original telegraph submarine cables were just dragged up the beach, buried a little, then then ends cut and wired up in a tiny concrete hut at the end of the beach. Do an image search for “Porthcurno cable hut” and read about it at https://en.wikipedia.org/wiki/Porthcurno
Probably. Washington State's cable landings are all in Puget Sound, a similar distance up the Strait of Juan de Fuca.
In the Pacific Northwest the weather on along the coast can get pretty fierce in the winter (wind, storms). You don't tend to see a lot of industrial infrastructure built there.
No. Once they reach land they use public right-aways just like normal service providers. There would be no reason to bury a cable in a river to get it from the coast to the city.
Well, technically they're still on the coastline. Just like docks, marinas, and even cities, they tend to get moved inland along harbors and estuaries where available, for protection from ocean weather and waters. Even San Francisco ('s city center) is on the Bay-facing side of the relevant landmass.
There's a really good book called 'Tubes" where the author describes visiting a cable landing site. Also being present when a cable is bought ashore, and when the cable laying ship is loaded.
Equinix used to offer this map as a cool poster to customers (and maybe everyone?). I'd love a not-beat-up replacement, but I don't see an order link like there used to be. I did find the 2019 one for $250(!!), which is a little more than I want to pay.
I figured there would be something like that. But then, why are the submarine cables not hidden as much ? Because they are physically harder to access ? Because they travel in shared maritime territory ?
Anyway, thanks for the read, the article and paper were interesting !
For some reason I was surprised to see so many cables not spanning oceans. Like all the ones going the Strait of Hormuz and up the the Suez Canal. But I guess it just makes a lot of sense to lay cable where you don’t have to worry about buildings and people being in the way.
Crossing multiple national borders is a major political undertaking.
It is far more simpler to lay a subsea cable in international waters, especially when hostile nations are concerned.
As to the Strait of Hormuz and the Suez Canal, there really are few or no alternatives, that's why there are so many cables there. It's also why Egypt makes serious bank on permits and other fees. We are talking hundreds of millions of dollars.
I'm surprised to see that there are so may cables just going along the coast. I guess it makes sense, because if you want to connect coastal destinations, by going through the ocean, you don't have to dig or install telephone poles, you just lay the cables down there.
No, utlity poles are MUCH, MUCH cheaper than undersea cables. Not even close. Plus you can run 216-strand on telephone poles whereas undersea cables (due to repeaters) are generally 6-strand and almost never more than 12-strand. Utility poles are cheaper even including the cost of the poles themselves.
However in order to put up utility poles you need a ROAD in order to get vehicles to the poles for maintenance. That's why you see offshore cables: when there are no roads through politically-stable territory.
Look at Alaska: the fiber runs overland all the way from the southern coast to the northern coast because the oil companies built a road next to the pipeline. Then it goes offshore along the arctic coast because there are no roads between the coastal communities.
Once you have roads and political stability, utility poles win every time. Underground is more reliable but more expensive. Undersea is insano-expensive.
Subsea cable system costs are about 30k$ to 50k$ per kilometer. This is about the same cost as aerial construction, excluding the poles. You pay for the poles separately with yearly fees.
You also don't need roads to put up and service utility poles. It's just more convenient to put them by roads.
As an aside, you can and do have subsea cables with 216 or other high count strand numbers. The reason you use 12 strands or less in long haul systems is that it is cheaper to use fewer strands and higher bandwidth rates per strand than to build high count subsea cable systems.
You could build any terrestrial fiber system with just one strand, but it's cheaper and more convenient to use more strands since you are constantly bifurcating the network.
Source: I have built both subsea cables and terrestrial fiber networks.
> This is about the same cost as aerial construction,
Not even close. 12-count ADSS cable costs $3,000 per km, one tenth of your low-end. And for three times that price you can get 144-count which isn't even remotely an option for subsea cables.
I used to buy from FS.com, but they've stopped shipping large spools due to the COVID shipping disaster situation, and took the pricing down off their website. Here's another company (never dealt with them) charging $1,400 per km:
> Show me a 216-count subsea cable long enough to need repeaters/amplifiers.
I know reading is hard, but repeat after me: "The reason you use 12 strands or less in long haul systems is that it is cheaper to use fewer strands and higher bandwidth rates per strand than to build high count subsea cable systems."
If you want high count subsea cable segments with amps, but which does not use in-line amps, take a look at https://crosslakefibre.ca/
To spell it out for you, subsea cables which are only a few hundred kilometers long do not require in-line amps, only head end amps.
I don’t understand what your problem is? You can read, can’t you?
Google even pops up a box with the relevant information immediately. Here, let me help you if that’s too hard:
“The cost of completing the nascent fiber-optic network connecting the capital cities of Sub-Saharan Africa and the main submarine cables is modest at $316 million, based on a cost of around $27,000 per kilometer.”
And that’s just the top result. You can look up the per km cost of almost any cable system trivially.
But I wonder how the line between Texas and Missouri got there.
And the one between California and Washington: https://www.submarinecablemap.com/#/submarine-cable/pacific-... (I get that it's to complete a ring, but if subsea is so expensive, they couldn't put together an in-land route to make it a ring?). Or is it cheaper to make your sea cable 5% longer instead of licensing land-RightOfWays install/access?
> But I wonder how the line between Texas and Missouri got there.
Would not be surprised if there were an offshore oil drilling platform in the middle of it.
> And the one between California and Washington:
Pretty sure that's just how the map is drawn -- the consortium that laid the subsea lines to Japan probably bought six strands of terrestrial fiber between the two US endpoints to make the ring.
This was my thought, particularly with the number of municipalities one would traverse by land, and all the varying local and state governnments one would encounter.
I think that’s why so many will follow a rail, pipeline or highway right of way. Often these are state or federally regulated, sometimes letting them do whatever they want. In Canada, the railways pre-date the country and have some interesting rights and privileges.
But if a subsea line is 6 fibres, you’d think they’d license one of the 288 strands on an existing run from some less else for less than that cost.
> But if a subsea line is 6 fibres, you’d think they’d license one of the 288 strands on an existing run from some less else for less than that cost.
This is easier said than done.
There may not be spare strands on the route in question, or if there is, it may not be available at a reasonable cost. Remember dark fiber is an unregulated asset, nobody can force network owners to play ball.
The available terrestrial routes between the landing stations may be suboptimal.
The route between landing stations may be required to be diverse from terrestrial route.
There's a whole another level of control between owning a full route and leasing some strands in somebody else's cable.
Well for Norway that proves the point. Have you saw the coast of Norway? It's heavily punctured by fjords which mean a lot of up and down. Many communities are connected to many of their neighbours by car ferries, not roads.
Hey iptrans, instead of just saying "octoberfranklin is wrong" over and over, how about posting a verifiable link with pricing information? You know, like I did here:
So I thought I'd try some long-haul traffic and was amazed that the MangaWhai Tavern[1] responds like it was next door. Those bits should be coming (at least when accessed for the first time in some cache-purge period) from NZ to Pago Pago to Kapalei HI, to Oregon and down the West Coast to me. I'd'a thought I'd see at least a little delay. Clearly the network is smarter than me.
> httping mangawhaitavern.co.nz
PING mangawhaitavern.co.nz:80 (/):
connected to 76.223.27.102:80 (201 bytes), seq=0 time=300.61 ms
connected to 13.248.155.104:80 (201 bytes), seq=1 time=288.97 ms
connected to 76.223.27.102:80 (201 bytes), seq=2 time=293.50 ms
Yep. We can get 4gbps home connections here now, yet general browsing experience can be slowish.
If you think about it, Huawei Marine and Huawei are not that related anymore (if I remember correctly Huawei has a very minority stake left on Huawei Marine).
It would nice if there was terrestrial fiber map which shows fiber lines in cities and towns. Also is there is a map available that shows where active fiber connections are available by address. FIOS doesn’t have a coverage map.
I found one a couple years ago that showed all the fiber lines on the streets and all that. I was using it to plan out talking Comcast into offering their 2gig.
That would be nice, but I haven't found any, even in other countries. Seems this is due to national security (if someone can break down the network, it will paralyze that country). But at least an idea of exactly where they are willing to install should exist (and does exist in other countries).
Heh, I once figured out where most/all the lines came into the city I was living in. Drove out there and was able to identify it: probably all the providers shared a group of fibre there. I'm sure it was a part of a ring, but fun to learn I knew where the internet came into town :)
Not really. The download speeds beta testers are seeing go up to 100mbps and that’s at low load, but the project isn’t aiming to replace submarine cables. The goal is to fix the problem of running fiber being uneconomical for most of the rural US (and elsewhere), so 100mbps is amazing speeds for end users who otherwise would be on slow DSL/satellite connections.
Adding to the above, Starlink doesn't have but a fraction of the bandwidth of a single wave in a single fiber optic strand.
Starlink just can't carry the bandwidth needed to connect the planet. The theoretical throughput of the whole Starlink constellation is equal to just a single subsea cable system.
And really his whole superb novel Cryptonomicon. But that’s 1,000 pages.