Every so often, a big chunk of the internet vanishes for an hour — a major site or even a whole country becomes unreachable — and the post-mortem blames three letters: BGP. It's one of the most important protocols on the internet and one of the least known. Let's fix that.
What BGP does
The internet isn't one network — it's tens of thousands of separate networks (ISPs, clouds, companies) stitched together. Each is called an Autonomous System (AS). BGP (Border Gateway Protocol) is how these networks tell each other which destinations they can reach, so traffic can find a path from any AS to any other.
In short: BGP is the internet's routing map. It's how a packet leaving your ISP knows which series of networks to hop through to reach a server on the other side of the world.
How it works (simply)
| Step | What happens |
|---|---|
| Advertise | Each network announces "I can reach these IP ranges" |
| Exchange | Networks share these announcements with their neighbours |
| Choose | Each network picks the best path to each destination |
| Route | Traffic follows those chosen paths, hop to hop |
It's a giant, constantly-updating conversation between networks about who can reach what — and the paths it produces are what traceroute reveals when you trace a route.
Why it's fragile
Here's the catch: BGP largely runs on trust. Historically, if a network claims it can reach a set of IPs, others tend to believe it. That's efficient — and dangerous.
| Failure mode | What happens |
|---|---|
| Route leak / misconfiguration | A network accidentally announces routes it shouldn't, black-holing traffic |
| BGP hijack | A network (maliciously or by error) claims someone else's IPs, diverting traffic |
| Withdrawal | A network stops announcing its routes — and effectively disappears |
The famous outages — where a big platform or whole region drops off the internet — are often a BGP misconfiguration: a network accidentally withdrew or mis-announced routes, and suddenly nobody could find it. The servers were fine; the map was wrong.
Improvements like RPKI (which cryptographically validates route ownership) are steadily making BGP safer, but its trust-based core is why these incidents still happen.
Do you need to deal with BGP?
For most people: no, but you should recognise its symptoms. You don't configure BGP unless you run network infrastructure. But when you see:
- A site unreachable from some regions but fine from others,
- A traceroute that dies or detours wildly at an early hop,
- "The server is up but unreachable" with no server-side fault,
…a routing/BGP issue upstream is a real possibility — and it's usually out of your hands (your ISP's or a transit provider's). Recognising it saves you from chasing a phantom server bug.
A monitoring angle
Because BGP problems are often regional, monitoring from multiple locations is how you spot them: "down from Europe, up from the US" is a classic routing-problem signature that a single-location check would completely miss.
The bottom line
| In one line | |
|---|---|
| What | The protocol that routes traffic between the world's networks. |
| Role | The internet's constantly-updating routing map. |
| Fragility | Trust-based — misconfigs and hijacks can black-hole traffic. |
| You | Recognise the regional symptoms; the fix is usually upstream. |
BGP is the invisible glue holding the internet's networks together — and when it slips, the effects are dramatic and global. You won't touch it day to day, but knowing it exists turns a baffling "half the world can't reach us" into a recognisable routing problem.
Related: ping, traceroute & MTR, anycast explained, why monitor globally.