How the Internet Connects Across Countries and Continents: Undersea Cables, Satellites and Routing

Key takeaways

• The internet is mostly physical. Glass fibre on the seabed, not satellites, carries nearly all traffic between continents.
• Light in fibre travels at about two-thirds of its speed in vacuum, so distance sets a minimum delay. Roughly 5 ms one way per 1,000 km.
• Networks (autonomous systems) join each other by peering at exchange points or by buying transit, and BGP is the protocol that tells them which paths exist.
• Much of what you load comes from a CDN server near you, so it never crosses an ocean.
• Cable cuts, mostly from ships' anchors and fishing, are common. Routes are redundant, but a cluster of cuts in one corridor still slows traffic.

What connects the internet across countries?

Four things do the work: submarine cables, landing stations, networks that interconnect by peering or transit, and (for special cases) satellites. Together they make the "network of networks" that we call the internet. No single organisation owns it.

Are undersea cables really the main route?

Yes. Industry and ITU sources commonly put the share of international data carried by submarine cables at around 99 percent. Satellites carry a small fraction because fibre offers far more capacity at lower delay and lower cost per bit. The exact number of cables in service and their total length change every year as new systems are laid, so check a live source such as the TeleGeography submarine cable map instead of trusting a fixed figure in any article.

How does a submarine cable work?

A submarine cable is a bundle of fibre pairs inside layers of protection. Near shore, where ships and fishing gear are a danger, it is armoured with steel wire and often buried in the seabed. In deep ocean it is thinner, because less can disturb it.

1. Lasers at the landing station send data as pulses of light down each fibre.
2. Light fades over distance, so repeaters (optical amplifiers) boost it roughly every 50 to 100 km along the route. They are powered from the shore by a copper conductor in the cable.
3. Modern systems use dense wavelength division multiplexing, which sends many colours (wavelengths) of light down one fibre at once, each carrying its own data stream.
4. At the far landing station the light is converted back into electrical signals and handed to a local network.

Capacity per cable is very large, in the hundreds of terabits per second for the newest designs, but it is shared among many operators. Do not read that as "your" bandwidth.

What happens when you open a website hosted abroad?

Here is a simplified path for a request from a laptop in Pune to a server in another country:

1. Your laptop asks DNS for the server's IP address. The answer may already be cached by your router or ISP.
2. Your packet goes to your home router, then to your ISP's network.
3. The ISP looks up the best route to the destination network. It learned its options through BGP.
4. If the destination is abroad, the packet travels over your ISP's backbone to a landing station in a city such as Mumbai or Chennai, and goes under the sea.
5. At the far end it enters the destination country's networks, possibly through an IXP, and reaches the server's network.
6. The reply makes the same journey back, often by a different path.

The outward and return paths can differ because each network decides its own routes. This is called asymmetric routing and it is normal.

How BGP decides the path

The internet is made of tens of thousands of autonomous systems, each an organisation's network with its own number (an ASN). They use the Border Gateway Protocol, defined in RFC 4271, to announce which IP ranges they can reach. BGP picks routes using policy as well as distance: price, business agreements and preferences count as well as the shortest path.

Networks connect in two main ways:

• Transit: a smaller network pays a larger one to carry its traffic to the rest of the internet.
• Peering: two networks agree to exchange traffic between themselves, usually free when both benefit. IXPs make peering cheap, because one connection reaches many networks.

Because BGP relies on trust between networks, mistakes and hijacks happen: a network announces ranges it should not, and traffic is misdirected or dropped. Filtering and RPKI (a system for cryptographically validating route origins) are the main defences.

See the path yourself with traceroute

You can watch this on your own computer. traceroute on Linux and macOS, or tracert on Windows, lists each router (hop) between you and a destination and the time to reach it.

traceroute example.com # Linux, macOS
tracert example.com # Windows

Read the output like this. Early hops are your router and ISP. A sudden jump in latency, for instance from 20 ms to 150 ms or more in one hop, usually marks a long-distance link such as an undersea cable. Router names often contain city or airport codes that hint at where the hop is. Some hops show * because the router does not reply to probes; this is not necessarily a fault.

A quick calculation shows why distance matters. A one-way path of 10,000 km through fibre needs about 50 ms for the light alone, so a round trip cannot be under 100 ms however fast the equipment is. Real paths are longer than straight lines, and routers add delay, so actual figures are higher.

What role do satellites play?

Satellites fill gaps that cables cannot. Ships, aircraft, islands and remote villages use them, and they act as emergency backup if a cable is cut.

• Geostationary satellites sit about 35,786 km above the equator. The round trip to space and back adds roughly half a second or more of delay, which is why calls on these links feel slow.
• Low Earth orbit constellations fly at a few hundred to a couple of thousand kilometres. Delay is much lower, in the tens of milliseconds, but each satellite covers a small area and capacity per user is limited.

Satellites complement cables; they do not replace them for the bulk of international traffic.

Why most content never crosses the ocean

Large sites and streaming services copy their content to Content Delivery Network (CDN) servers inside ISP networks and at IXPs close to users. When you watch a video or load a popular site, you are often served from a server in your own city. That is why a site hosted abroad can feel fast. Content that cannot be cached, such as a live database query, still has to travel to its origin.

Risks to global connectivity

• Accidental damage. Ship anchors and fishing gear cause most cable faults. Repair ships have to be sent out, which can take days to weeks.
• Natural events. Earthquakes and underwater landslides can break several cables at once.
• Geopolitical tension. Cables in contested waters, such as parts of the Red Sea, have been affected in recent years, slowing traffic between Asia and Europe.
• Concentration. Many cables share the same corridors and landing cities. Redundancy exists, but not infinite diversity.
• Routing errors and hijacks. Not physical, but they can take whole services offline.

Our post on threats to undersea cables covers the security side in more depth.

Why this matters for a career in IT

If you work in networking, cloud or security, you will meet these ideas daily. Choosing a cloud region is partly a latency decision. Diagnosing slow applications means reading traceroutes. Securing an organisation means understanding BGP and DNS, which sit at the heart of how traffic finds its way.

Next steps

Run traceroute to three sites in different countries and note where the latency jumps. To go deeper, read how a computer network works and what ISPs do. For structured learning of routing, BGP basics and troubleshooting, see the CCNA 200-301 course.