Technology & Computing · Internet & networks
How the internet works
The internet is tens of thousands of separate networks that agree to carry each other's traffic. See how data is cut into packets, how routers pass them along, and how they're put back together.

What you will learn
- Describe the internet as many separately run networks that agree to connect
- Explain how networks reach each other through transit, peering and exchange points
- Explain why data is cut into small addressed packets instead of sent down a reserved line
- Describe how routers pass each packet along, one hop at a time, using its destination address
Lesson 1 of 4 · 4 min
Nobody owns the internet, so who carries your data?
Send a photo to a friend abroad and it crosses equipment run by several different organizations. No single company owns the whole route, and none could.
The internet is a network of networks. Your home network, your provider, a university and a streaming company each run their own network, with their own rules.
A network that runs its own routing this way is called an autonomous system, and each one gets an ID number. More than 100,000 of those numbers had been handed out by 2020.
No central body runs the whole thing. The networks connect by agreement, and they work together because every one of them follows the same set of rules, the internet protocols.
Your home router sends anything not meant for your own devices to your internet provider. Your provider can't wire itself to every other network. So it buys transit, paying a bigger network to carry its traffic to the rest of the internet.
Networks of similar size often agree to peering instead. They swap traffic for free, because each gains about as much as it gives. Tier 1 networks sit at the top. They reach every other network through peering and pay no one for transit.
Many of these handovers happen at exchange points, data centers full of switches where many networks plug in side by side. Most are run as not-for-profit associations of their members.
Since about 2010, companies such as Google, Netflix and Meta have connected straight to providers. That lets their traffic skip the top tier.
Try to recallWhen two internet networks connect, what's the difference between transit and peering?
With transit, a smaller network pays a bigger one to carry its traffic to the rest of the internet. With peering, two networks of similar size swap traffic for free, because each gains about as much as it gives.
The internet is thousands of separately run networks with no one in charge. You can now say how your data crosses them, through paid transit, free peering and exchange points.
Lessons in this topic
Nobody owns the internet, so who carries your data?
See the internet as thousands of separately run networks, and how they pay or swap with each other to reach everyone.
Why your data travels in thousands of small pieces
Learn why the internet cuts data into addressed packets instead of reserving a line, and how routers pass each one along a hop at a time.
How IP and TCP split the job of delivery
See why the internet's basic delivery rules promise almost nothing, how TCP turns that into a perfect copy, and why the split has lasted since the 1970s.
How one wrong message can knock a site off the internet
See how networks use BGP to tell each other what they can reach, and how one bad announcement cut off YouTube in 2008 and a missing one cut off Facebook in 2021.
Three quick stories from this topic
Invented twice
Packet switching was invented twice, separately, in the United States and Britain
Paul Baran worked on it at the RAND Corporation in the early 1960s, wanting a network that could survive an attack. Donald Davies at Britain's National Physical Laboratory reached the same idea in 1965.
The ideas meet
Davies's team presented their work, citing Baran's, at a 1967 symposium and persuaded the ARPANET's planners to use packets.
Doubted at first
Telephone engineers argued that router buffers would quickly run out, and later that packet networks couldn't pay without government money.
Internet by pigeon
A Norwegian group sent nine internet packets by pigeon in 2001 and got four replies
They followed RFC 1149, a 1990 April Fools' standard that prints each packet on a scroll taped to a bird's leg. Over about 3 kilometers, replies took between about 50 and 100 minutes.
Why it works at all
IP promises only best-effort delivery, so any link that can carry a packet will do. The joke standard said a lost pigeon was acceptable.
Still a joke
RFC 1149 calls itself experimental and not recommended. Its author added a quality-of-service version in 1999.
The two-napkin protocol
The protocol that steers traffic between the internet's networks began on two napkins
At an engineering meeting in Austin, Texas, in January 1989, Yakov Rekhter, Len Bosack and Kirk Lougheed sat at a table and sketched the Border Gateway Protocol. The design grew to three handwritten sheets.
Still in charge
BGP still carries routes between networks. A full table of internet routes now holds more than a million address blocks.
A copy on the wall
A photocopy of the three sheets hangs in a routing development area at Cisco in Milpitas, California.
Sources
- Internet, Wikipedia
- Autonomous system (Internet), Wikipedia
- Understanding How Facebook Disappeared from the Internet, Cloudflare blog
- Router (computing), Wikipedia
- Tier 1 network, Wikipedia
- Internet exchange point, Wikipedia
- Packet switching, Wikipedia
- IP over Avian Carriers, Wikipedia
- RFC 1149: A Standard for the Transmission of IP Datagrams on Avian Carriers, IETF RFC Editor
- Yakov Rekhter, Wikipedia
- Border Gateway Protocol, Wikipedia




