Copper, Fiber, and the Limits of Distance
Copper Ethernet has a hard limit you can almost feel: about a hundred meters. Past that, the signal degrades, the noise wins, and the link becomes unreliable. If you need to go further, you have to install a repeater, a switch, or a different medium entirely.
This is not a software limit. This is a physics limit. You are pushing electrons down a thin piece of metal, and as you push them further, the signal smears, attenuates, and picks up noise. There is no clever algorithm that fixes this. You either shorten the run, amplify the signal, or change the medium.
So we changed the medium. Most long-distance data today travels not on copper but on glass.
Fiber optics — a thread of glass and a flash of light.
A fiber optic cable is, at its core, an extraordinarily pure thread of glass — a core — wrapped in another layer of glass — cladding — with a slightly different refractive index. When you fire a light pulse down the core, the difference in refractive index causes the light to bounce off the inner wall of the cladding instead of escaping. The pulse propagates down the cable by total internal reflection. It does not "travel through" the fiber so much as it ricochets down it, contained.
The implications are dramatic.
A standard piece of fiber optic cable can carry a clean signal for around 100 kilometers without an active repeater. Premium fiber can carry it further. Compare that to 100 meters of copper. The medium is roughly a thousand times more capable on raw distance.
Fiber also draws far less power, because you are pushing photons through glass instead of electrons through resistance-heavy metal. It is immune to electromagnetic interference, because light does not care about the electric fields of nearby motors and ballasts. And it carries dramatically more bandwidth per strand, because the carrier frequency of light is enormous compared to the modulation rates of copper.
So why is your desk still wired with copper?
Because copper is cheap, easy to terminate, and backwards-compatible with thirty years of installed hardware. To bring fiber to a desk, you need transceivers at both ends to convert electrical pulses to light and back — modern modules called SFPs or QSFPs, which are not free, especially in volume. You need carefully cleaved fiber endings and clean connectors, because a speck of dust on a fiber face can ruin a link. You need network equipment that supports fiber ports.
In an office of two hundred desks, the math almost always favors copper to the desk and fiber between the floors or buildings. ROI rules everything around us.
But step out of the office. The internet you are using right now is, at its long-distance backbone, almost entirely fiber.
The undersea cables.
When you ping a server in the US from India, you are not bouncing off a satellite. You are sending light through a glass thread on the floor of an ocean.
Look up the Submarine Cable Map (it is a real, open dataset). There are hundreds of fiber optic cables laid across the floor of every ocean, connecting every continent. These cables are how the internet actually exists between countries. Satellites carry a small fraction. The rest is glass on the seabed.
Each cable is laid by specialized ships, repaired by other specialized ships when they break (which happens, often from fishing trawlers and anchors), and is owned by consortia of telecoms and cloud providers. The cost to lay a single transoceanic cable is in the hundreds of millions of dollars. The latency of the global internet is, in a very real sense, a function of where these cables physically run.
This matters more than most developers realize.
If your servers are in Mumbai and your users are in São Paulo, your packets are taking a multi-hop fiber journey across continents. There is no software optimization that beats geography. The fastest you can possibly serve that user is the speed of light through glass between those two points, plus the time spent on every router in between.
Quick: how fast is the speed of light in fiber?
Light in vacuum travels at about 300,000 km/s. Light in fiber travels at about two-thirds of that — roughly 200,000 km/s — because it is moving through glass, not vacuum.
So India to the United States, in a straight line through the Earth — let's call that around 12,000 km — would be:
12,000 km ÷ 200,000 km/s ≈ 60 ms
Theoretical one-way latency. Round trip: about 120 ms. That is the physics floor. You cannot do better with fiber.
Practical ping times from India to a US server are usually 150 ms to 250 ms, and sometimes more. That is more than the floor. Where does the extra time go?
That is the next chapter.
Push On It
- Find the Submarine Cable Map online. Identify two cables that connect your country to another continent. Note their landing stations. How does that geography line up with where the major cloud regions are?
- From your laptop, traceroute a server on another continent. Count the hops. Note where in the world each hop appears to live (you can use any IP geolocation tool).
- Estimate the theoretical minimum one-way latency between your current location and a US data center. Compare to your actual measured ping. Where did the rest of the time go? List the candidates.