Why the Internet Goes Down: The Science of Service Outages

— ny_wk

A service outage is what happens when the invisible machinery of the internet stumbles — and almost every one of them traces back to a handful of surprisingly fragile points: misconfigured routing, severed cables, overloaded servers, or a single fat-fingered command. Understanding why the internet goes down reveals just how delicate the world's most important network really is, and why a glitch in one data center can darken half a planet.

We treat connectivity like running water — always on, always there. Then one morning a banking app spins forever, a streaming service throws an error, and an entire country's news feeds fall silent. That moment of digital darkness is a service outage, and behind every one of them lies a chain of cause and effect that is part engineering, part economics, and occasionally part politics.

What Actually Happens During a Service Outage

The internet is not one machine. It is roughly 70,000 independent networks — called autonomous systems — stitched together by agreements to pass each other's traffic. When you load a page, your request hops across several of these networks before reaching a server and bouncing back. A service outage occurs when any critical link in that chain breaks.

Engineers sort outages into a few recurring families. Knowing them is the difference between panic and a calm "ah, it's probably DNS."

The unsettling truth is that the most catastrophic outages rarely come from hackers. They come from routine maintenance gone wrong — a single typo in a configuration file, pushed to thousands of machines in seconds.

BGP and DNS: The Two Quiet Giants That Can Topple the Web

If the internet has a nervous system, it is the Border Gateway Protocol, or BGP. This is the protocol networks use to tell each other "send traffic for these addresses to me." It runs almost entirely on trust — and that trust is its weakness.

In October 2021, Facebook, Instagram, and WhatsApp vanished from the internet for roughly six hours. The cause was not an attack. A routine maintenance command accidentally withdrew the BGP routes that told the world where Facebook lived. To the rest of the internet, the company simply ceased to exist. Worse, the failure locked engineers out of their own remote tools, forcing staff to physically access data centers to fix it.

The second giant is the Domain Name System (DNS) — the internet's phone book, translating factfactory.com into the numeric address a computer actually dials. When DNS fails, servers are alive and well but unreachable, because nobody can look up their address. A botched DNS update was at the heart of the massive 2016 Dyn outage that knocked out Twitter, Netflix, Reddit, and Spotify across the eastern United States.

Both systems share a chilling property: they are globally interdependent. A mistake in one operations center can ripple outward in seconds, taking down services that have no direct connection to the company that made the error.

The Cables Under the Sea — and the Switches in Government Hands

For all our talk of "the cloud," about 99% of intercontinental internet traffic travels through submarine fiber-optic cables lying on the ocean floor. These bundles, often no thicker than a garden hose, carry the data of entire continents. When a ship's anchor drags across one, or an undersea landslide snaps it, whole regions can slow to a crawl.

Coastal and island nations feel this most sharply. Countries served by only one or two cables can lose the bulk of their connectivity from a single accident hundreds of miles offshore. Repair ships can take days or weeks to reach the break and splice it back together in the open sea.

Then there is the most deliberate outage of all: the government-ordered shutdown. Authorities can compel local internet providers to throttle or fully switch off access, often during elections, protests, or exam periods to curb cheating. Bangladesh, India, Iran, and others have all imposed nationwide or regional blackouts. These are not failures of technology at all — they are the network working exactly as instructed, just instructed to go dark. Digital-rights groups track hundreds of such shutdowns worldwide every year, each one cutting millions of people off from news, banking, and emergency information.

How Engineers Fight Back Against Downtime

The defining metric of online reliability is the "number of nines." A service with "five nines" of uptime — 99.999% — is allowed barely five minutes of downtime per year. Achieving that is a relentless engineering discipline.

• Redundancy: Run everything at least twice, in separate locations, so one failure never takes the whole system with it.
• Failover: Automatically reroute traffic to healthy servers the instant a problem is detected.
• Content Delivery Networks (CDNs): Cache copies of websites in hundreds of cities so a single origin server's death isn't fatal.
• Staged rollouts: Push changes to a tiny slice of users first, catching the typo before it reaches everyone.
• Chaos engineering: Deliberately break parts of the system in tests — famously pioneered with a tool that randomly kills live servers — to prove the rest survives.

Despite all of this, perfection is impossible. The internet was designed to survive damage, not to never break. Every outage that flickers across your screen is, in a strange way, a reminder of how astonishingly often this sprawling, improvised, trust-based machine actually works.

5 Mind-Blowing Takeaways

• Most major internet outages are caused by human configuration errors, not hackers — a single mistyped command can erase a company from the web.
• The 2021 Facebook blackout was a BGP mistake so severe that engineers were locked out of their own buildings and tools.
• Roughly 99% of international internet traffic runs through undersea cables thinner than a garden hose.
• "Five nines" reliability allows just about five minutes of downtime per year — the gold standard for critical services.
• Some outages are intentional: governments order hundreds of internet shutdowns worldwide every year, switching off access on purpose.

Frequently Asked Questions

Why do whole apps go down at once instead of just slowing down?

Because modern apps depend on shared backbone systems like DNS, BGP routing, and centralized cloud regions. When one of those foundational layers fails, every service riding on it disappears simultaneously rather than degrading gracefully.

Is an outage the same as a hack?

Usually not. Most outages are accidental — a bad software update, an expired certificate, or a cut cable. A deliberate attack, such as a distributed denial-of-service (DDoS) flood, is just one of many possible causes, and statistically not the most common.

What can I do when a service I rely on is down?

Check an independent status tracker or the provider's official status page, since the outage may be on their end, not yours. Restarting your router rarely helps with a large-scale provider failure — the fix is almost always out of your hands, and patience is the realistic strategy.

Why are some countries hit harder by outages than others?

Nations connected by only one or two undersea cables, or with few competing providers, have less redundancy. A single cable break or one government order can knock out connectivity for an entire population, while richer, denser networks simply reroute around the damage.

The next time your screen freezes mid-stream, you'll know there's a hidden world of cables, protocols, and split-second decisions fighting to bring it back — follow The Fact Factory for more of the science hiding in plain sight.

🤯 Love facts that rewire your brain? The Fact Factory drops a new one every single day.

• 📺 YouTube: @factsandstoriestube — subscribe for daily fact shorts
• 📸 Instagram: @factfactory57
• 📘 Facebook: The Fact Factory