How do they maintain them?
At what point to these become considered "national security" assets (in the eyes of the owning company's nation)?
Do they rent them out as business as well?
(This blog looks like it could have many such answers, but looking for a HN-comment-sized answer.)
Aren't most of them likely to be privately owned?
Microsoft, Amazon, and Google are known to have undersea cables. I bet there are a bunch owned by companies that aren't generally known to regular people. For example, Tata, which AFAIK has the largest amount of undersea cables.
Also, international waters outside of the boundaries of countries aren't owned by anyone. They probably should be owned by UN to help with problems like overfishing.
What's probably more interesting is what law governs them, and it turns out they're regulated under several conventions, including a dedicated one from 1884 and the UNCLOS (United Nations Convention on the Law of the Sea).
[1] has a pretty good summary if you're curious.
[1] https://www.noaa.gov/general-counsel/gc-international-sectio...
Quite a few, mostly run by telecom companies, often with a few big users/governments in partnership (including Google/Meta/Amazon etc). A Meta supported Atlantic cable was completed last year.
As big tech has been concentrating and also doing more global networking, and running their own backbones and things, they became heavy users of these cables, and then partners, and now sometimes sole owners.
I suspect the actual maritime operations are contracted out.
I don't know how accurate this is, but it seems like a good start towards a list of cables where big tech is involved [1]
[1] https://blog.telegeography.com/telegeographys-content-provid...
Decades ago.
Some other country decides to pull shenanigans and cut / hack the cable, in which case the company will insist the U.S. gov't get involved, maybe pay for the repair. Saying it's crucial for national security.
Microsoft/Google/AWS rent out this capacity in the form of networking capacity on their respective clouds. Your bits between their DCs likely doesn't traverse the public Internet (because that costs them more).
https://subseacables.blogspot.com/2024/10/facebooks-semi-sec...
https://www.theverge.com/c/24070570/internet-cables-undersea...
who ever thought this was a good idea?
I was thinking: "They really should add two toggles to the top of the page: one which makes the images static so that the scroll-problem disappears and one where all the non-relevant, non-technical information is hidden."
[0] mirrors: https://archive.is/IpfNq / https://archive.ph/IpfNq / https://archive.vn/IpfNq / https://archive.md/IpfNq / https://archive.fo/IpfNq
https://euripides.dk/setebos/frx/matrix/ai/books/stephenson_...
Stephenson discussed this a bit in the Long Now launch event[0] for "Polostan" and more or less confirmed what you suspect. He also went on to say that he has learned to reel that in and avoid those "rabbit holes". He avoids such digressions in Polostan, in my opinion to the detriment of the book.
> A Thread Across The Ocean by John Steele Gordon
A Web Around the World, Part 1: Signals Down a Wire January 7, 2022 https://www.filfre.net/2022/01/a-web-around-the-world-part-1...
What governments do you have to go to to get approval to do this? Could I just run a string across the Atlantic Ocean?
If we do lay cables on top of other cables how high do they get stacked? Are there challenges to bring the lower cables back up? Does that happen? Or do we just keep them down there forever basically and upgrade the hardware at the terminal?
I don't know exactly how often this has occurred but I'd guess it's relatively rare. The companies that operate in this space are very specialized and sophisticated. The locations of pretty much every cable laid in the last half century is very precisely tracked and one of the first things that has to happen when preparing a new cable route is to undertake a high resolution side-scan sonar survey of all or part of the planned route. In shallower water the cables are typically buried under several meters of the seabed.
> What governments do you have to go to to get approval to do this? Could I just run a string across the Atlantic Ocean?
At the very least you'll need to have landing agreements with the countries at the various endpoints. In international waters I believe there are some laws that apply but I gather that it's more about liability. You'd have a lot of difficulty running a string across the Atlantic. Controlling the amount of slack on a cable that's being played out is incredibly finicky work. Keep in mind that the point where your hypothetical string is touching down on the sea bed might be several miles behind where you are and that your ship is going to be bobbing around on the surface and you get an idea.
> If we do lay cables on top of other cables how high do they get stacked? Are there challenges to bring the lower cables back up? Does that happen? Or do we just keep them down there forever basically and upgrade the hardware at the terminal?
Cables are routinely brought up for repair or disposal. The ships that do this are called Agreement ships. In 1866 the second-ever transatlantic cable was grappled up to the surface and repaired (it snapped while laying it the previous year).
Modern cables are fiberoptic and do not increase their bandwidth once laid.
When the cables aren't well buried, they can migrate. No link, but earlier today I saw a description of a repair that was significantly delayed because the cable ends were 15 km away from where they were expected to be.
That said, I think there's a lot of area on the sea, and not a lot of cables, so chances of overlapping are low; especially if a survey is done of the new route immediately before. Although if they're buried, maybe you can't see them so well. And a little overlap here and there probably isn't a big deal, because most of the time cables are brought up, it's because they were severed, so likely it doesn't disturb the other cable too much on its way up.
No idea if specific care is taken to avoid that, or if it's even a problem, but I could imagine so for cables that are buried in the seabed – the act of burying the new one could damage the existing one.
But cables are only buried in shallow waters, is my understanding (where there's high risk of an anchor destroying them), so as long as the point of overlap or intersection is in deep seas, I'd imagine it to not be a problem.
The other reason that overlapping would be an issue is if they need to pull up a section to repair it and it's directly under another cable.
But the ocean is really damn big and there are only O(100s) of submarine cables. In both the burying in shallow water case, and the deep water maintenance case, I'm guessing the chances of two cables being in the same place is very very small.
No kind of expert at all, but I understood that better control over / perception of narrower bandwidths of light have allowed fiber-optic cables to improve their data throughput immensely. Is that incorrect? Or are you using a narrower, technical definition of "bandwidth" that I've not understood?
For short-ish cables (a couple hundred kilometers), driving the amplifiers via light (interestingly this can be done from both the data-sending and receiving end!) is possible, but not for trans-oceanic distances.
If I'm reading this right, BPSK - used on the longest transpacific cables - can't really be upgraded to have higher bandwidth but QPSK and other modulation schemes used for shorter distances can be (?).
Huh, I had no idea it needs to be done so frequently.
> These days optical cable repeaters are photon amplifiers that operate at full gain at the bottom of the ocean for an anticipated service life of 25 years.
> Repeaters are a significant cost component of the total cable cost, and there is a compromise between a ‘close’ spacing of repeaters, every 60km or so, or stretching the inter-repeater distance to 100km and making significant savings in the number of repeaters in the system. On balance it is the case that the more you are prepared to spend on the cable system the higher the cable carrying capacity.
https://blog.apnic.net/2020/02/12/at-the-bottom-of-the-sea-a...
Optical amplifiers simply boost the signal. A photo is absorbed and then re-emitted with more power, but the same frequency and phase.
There is no optical/electrical/optical conversion.
A large factor for the efficiency, is how “thin” you can make channels in the cable via frequency of the light. Multiple signals can be sent over a single fiber line if they use different frequencies. One system I was working on at the time was constructing 88 channels at 100gbs for a total throughput of 8.8tbs.
Another very interesting aspect, is that service providers might share a lease of the same underwater cable, not actually owning the cable themselves. The companies enter an agreement on the frequency ranges assigned to them. Any mess ups in settings can cause disruption in your neighbor on the cable, resulting in millions of dollars of fines.
The levels of buying internet service simplified:
1. Pay for best effort traffic level. E.g. A residential user paying for 1gbps service
2. Pay for guaranteed traffic level. E.g. a business paying for guaranteed 10gbps. Often this maps to paying for a dedicated frequency on a cable, but that's hidden from you, you are just provided an Ethernet port or two.
3. Pay for a dedicated frequency on a cable. Also called "lease a wave". As you noted this is the most common way to go undersea. But it's common over terrestrial cables too. In this case the purchaser is often responsible for their own optical networking equipment, and the service provider muxes together all their customers. Sounds like you know more about this than me. The service provider's equipment doesn't have protection if your neighbor is accidentally stepping on your frequency range?
4. Pay for an entire cable. Also called leasing dark fiber. It's 100% your responsibility to "light" it. To make it worth it you're usually using optical networking equipment to mux together several different waves, just like the above scenario undersea, just all different waves belong to you. Commonly used by big tech companies and ISPs to connect their datacenters and points if presences in big cities. Lots of times the company you lease the cable from is also your contractor to repair cuts and maintain the amplifiers every ~50km.
You've got 3 types of use cases merged in one there.
First, The generic business internet which gives you uncontended bandwidth usage as far as an internet exchange point and an ISP which ensures its peering and transit is not oversubscribed.
Second, point to point ethernet connectivity. This morning I had an issue with a point-to-point connection I have from Beijing to the UK for example, where latency had jumped overnight from 215ms to 430ms. This is provided as point to point ethernet by I would assume something like MPLS or VXLAN over the providers network. I've had SDH and ATM backed ethernet services in the past too.
Third, I have link with an optical service from, which does map to a dedicated frequency as in your 3rd option, although it's provided to me as a standard ethernet handoff.
This is different to "pay for a dedicated frequency on a cable", as the service I have has an ethernet handoff. I couldn't put non-ethernet traffic on the frequency.
I've also got dedicated cables with a service provision at an ethernet layer (the provider gives me an ADVA which they control). The actual light goes from the ADVA in my equipment box to another ADVA in another equipment box
In none of these situations do I need to worry about frequencies, high power SFPs, etc, the handoff from the provider is always ethernet, they handle that.
My company does have some leased fibre between campuses too, I tend not to get involved in that, but I know one department runs some 800G sfps over a leased fibre between cities about 200 miles long. I don't think we have any owned fibre other than between buildings on the same campus
https://m.youtube.com/watch?v=VKpAIx0zG-U
https://nlnog.net/static/nlnogday2024/NLNOG2024-02-Yurii_Pol...
That's not true: The amplifiers they use work at the analog (and in fact even optical, i.e. without conversion to electric and back) level, and it's possible to upgrade capacity by only modifying the endpoints, not the entire cable.
I thought that endpoint upgrading of submarine cables was a sort of well known thing (at least amongst people that are aware their data goes and partly becomes a bunch of photons under the ocean, which I guess may be a lot of people here)
- The author.
It's a good question, the ocean being international waters and all.
I dont know that answer but for [geosynchronous] satellites in orbit this is negotiated with affected countries via the ITU.
(Create a mesh of undersea cables dense enough and you end up with a tripwire for submarines...)
- The Author.
Did you mean workplace? That's discontinued now.
So that mostly leaves “AI generated nonsense content” for Facebook and Instagram which surely people are going to get sick of fast…?
Just really not actually seeing where they actually get a useful product out of AI here…
Your comment made me think... What if the idea is that Facebook (I refuse to call it meta) is afraid of the very same thing you outlined and therefore thinks that better to be in control of whatever does the cannibalizing with summaries as opposed to letting the OS vendor or potentially web browser vendors control the "AI"?
I don't know the future but maybe it is possible that what we have now with AI slop is something like an uncanny valley and there is something better once we wade through this trash? Even if not, it is probably better for them to throw cash at this and it turns out to be a nothing burger than they miss it and now their whole business is in jeopardy.
Where did you read that? It’s certainly not in TFA. There’s a second article from the same substack linked on the page that speculates it “ may be heavily influenced by AI considerations”, which is not really the same as “potentially for AI”. https://subseacables.blogspot.com/2024/10/facebooks-semi-sec...
So it kind of is worth getting hung up about. But hey, today you learned this! :)
I would understand better if it was for VR, or simply video/communication.
But they still need fibre optic connnections to move immense amounts of data.
I feel like governments should be funding healthcare, housing and connectivity instead of missiles thousands of miles away from their borders.
The standard economic answer is that governments should at most finance public goods. But nothing of what you mentioned is a public good. (Well lobbing missiles at foreigners perhaps is a public good, but not one that's necessarily worth funding. Just like a light-house in the middle of nowhere might not be worth funding.)
1. Herd immunity and clean air from that Wikipedia link fall under public goods. Both of those are a subset of healthcare.
2. Basic human needs: many people would argue healthcare, housing, and connectivity should be universally accessible as a moral imperative
3. Broader social benefits. If people have these 3 things, they can be more productive.
4. "bread and circuses" is drawing a false equivalence between basic needs and entertainment/luxury goods. That's just bad argumentation.
That comment specifically listed "connectivity" as something he thought governments should fund, and only complained about governments funding weapons.
Governments can give money to poor people. Poor people can use money to buy goods and services.
(We don't need the government to provide housing etc to rich people. Rich people can help themselves, if we let them.)
What should we call such a thing though? This interconnected network of computers. How about the Inter-Conn-Net-of-Computers. Maybe someone could help me shorten that name.
(You can also buy stuff over the Internet these days, sure. And you can buy internet access with money, too. Isn't money wonderful?)
How do goods get to the store for purchase? Does any of that need the Internet? Like when the merchant pays for goods using a credit card?
For better or worse the Internet is critical infrastructure for modern society these days, it's not just scrolling on TikTok.
I suggested to give money to poor people. Not weird stamps and a list of approved goods in some approved store.
But yes, just like you can use money to buy food (a critical component of staying alive), rich or poor alike can also use money to buy internet access.
I'm not sure what your point is?
Standard economics are a simplistic model that fails hard to account for the effects of their own simplifications. No one sees physicists saying that a bouncy ball will bounce forever because it loses a negligible amount of energy in each bounce, but we see economists talk about a magical system where they get pretty much every assumption wrong and don't account for how it makes the model deviate from reality. (I had a list of aspects where the assumptions go wrong, but it's a bit too long to fit in the margin of this post)
Things like the Gini coefficient is over 100 years old and it showed that the economy could be doing great based on averages, despite it only being true for a few, yet there's still no push away from metrics that misrepresent how most people are doing.
Well, maybe they are also ignored scientists that we all know have ideas on how to improve things, but are powerless too.
Things obviously going downhill for 40+ years, but maybe working as intended, - https://en.wikipedia.org/wiki/Economic_inequality
> Things obviously going downhill for 40+ years, [...]
Global economy inequality has decreased markedly over the last few decades. More and more people have it better than ever before.
If housing price increase outpaces wage growth (which it does), with GDP growth being concentrated in very few pockets, this leads to decreasing living standards for most of the population, at least in this aspect.
Sure we have developers in Texas, Oklahoma, Florida, etc plopping down houses >$300-500k, but can the average person afford them? Yes I agree building codes in places like California make houses expensive, but the current market in the United States does not incentivize building affordable housing.
When more supply goes on the market, existing stock goes lower in price - when there is enough supply.
That is how it always has been.
Even old ‘shotgun shacks’ are too expensive because zoning restrictions (and too much easy money) has caused everyone to inflate everything.
> When more supply goes on the market, existing stock goes lower in price - when there is enough supply.
Yes, that's part of why the housing stock today is nicer than the housing stock 200 years ago: over time the average newly built home was always better than existing houses.
These are all areas with similar levels of ‘desirable is constrained’. Or perhaps worded better as ‘people want the top x percent, not just the same thing in bulk’.
Most of the other things you’re talking about, it’s easier to scale up production without hurting desirability.
For instance, if commute/neighborhoods/climate literally didn’t matter, everyone complaining about housing costs would just move to Rural Kansas or North Dakota and problem solved for dirt cheap.
Instead everyone is complaining about how nice new housing in city-of-choice-close-to-work is unaffordable.
Well of course it is - people are restricting building to keep it nice (by their standards), which is why folks want to move there, which is restricting supply - and with easier money (historically) that is causing increased bidding and increased costs.
Because the building codes drive up the costs!
However, fundamentally, even if fibre took the most direct route from your house, directly straight-line to the datacentre with the server in, and then straightline from there to your friends on the East Coast, the time taken to complete that journey and back is still going to be 150-200msec or so; so it won't be as snappy as if you all lived nearby, sadly.
So that's a theoretical minimum of 75 milliseconds additional lag in addition whatever network equipment exists in between you and your friends. Actually not too bad, even though we concede that the premise is slightly absurd.
This could potentially be 50ms latency saving from East Coast to SEA.
Unfortunately the tech is so far away it likely won't happen for another 10 years.
https://www.linkedin.com/pulse/map-metas-w-cable-sunil-tagar...
Are you saying the article is false?
Or that the map illustration is false? Although the map illustration doesn't come from your post.
How did you even find the LinkedIn post? Are they the same author? Is TFA based on the LinkedIn post? How do you know?
And it seems like the TFA doesn't even have an author, nor can I find an author for their whole blog...
Maybe you can clarify all of this, since you seem to have some context here?
“Roderick Beck worked as a sales contractor for Hibernia Atlantic and helps buyers procure capacity and providers make sales.”
AFAICT this post isn't "false" as much as "speculation". Given the news that a cable will be spanning the Atlantic ocean with an end goal of South Asia, South Carolina -> Africa doesn't seem insane. Though it looks like there's no cables there right now...
I was coming in here to complain about "encompass" vs "encircle", but now I'm fascinated by this map. Cool webdev, too!
EDIT: My biggest takeaway is that we should conquer/buy/steal French Polynesia. Also, huge shoutout to the Leif Erikson cable, connecting Oslo with the absolute middle of nowhere[1] in Canada. Oil rig thing, maybe...?
As noted the cable connects the US East Coast to South Africa and then heads to India and continues on to Australia before the home stretch to the States. We even know the number of fibre pairs, 16. It is a spatial division multiplexing system.
How long till we get Niger and Algeria connected?
I am the author. The article is not speculation. I know from sources inside Facebook that the cable will head Southeast from the US East Coast directly to South Africa. From there it heads directly to India. From India to Australia. From Australia to the US West.
What is speculation are the branching units. It is natural to add branching units to aggregate traffic to more countries. But I have no confirmation.
Regards,
Roderick.