Routing: how packets find their way

In lecture 1, packets crossed the internet “hop by hop”. That phrase hides one of networking’s most elegant ideas: no router knows the full path. Each one makes a single local decision — which neighbor should get this packet next? — and trusts the next router to do the same.

The routing table

Every router, and in fact every computer, keeps a routing table: a list of destination networks and where to send traffic for each. A typical entry has:

Destination

A network prefix such as 172.20.8.0/24. The special prefix 0.0.0.0/0 matches every address and is called the default route.

Next hop

The neighboring router to hand the packet to — or “directly connected” when the destination network is attached to this router.

Interface

Which port or link to send the packet out of.

Metric

A cost used to choose between routes to the same destination (for example, fewer hops or faster links).

You can see your own machine’s table with ip route on Linux, netstat -rn on macOS or route print on Windows. On a laptop it is usually short: the local subnet, and a default route pointing at the router.

Longest prefix wins

Often several entries match the same destination. A packet for 172.20.8.10 matches 172.20.8.0/24, 172.20.0.0/16 and the default route 0.0.0.0/0. The rule is simple: use the most specific match — the longest prefix. A /24 fixes 24 bits of the address, so it beats a /16, which beats the default /0.

This rule lets networks describe the world compactly. A router can say “everything in 10.0.0.0/8 goes left” and then add a single exception for one subnet, without listing millions of addresses.

Follow a packet across four routers

Step through the model. Watch the table of the router currently holding the packet, which rows match, which one wins, and how the TTL falls at each hop. In the last step a link fails and traffic takes another path.

TTL: protection against loops

Misconfigurations happen: router A sends to B, B sends back to A. Without protection, a packet would circle forever, consuming bandwidth. The Time To Live field prevents that. Every router lowers it by one; at zero, the packet is discarded, and the router usually sends an ICMP Time Exceeded message back to the source.

traceroute turns this into a diagnostic tool. It sends packets with TTL 1, then 2, then 3, and so on. Each router along the path discards the packet when its TTL runs out and reports back — revealing itself, hop by hop.

Where routes come from

Routers learn routes in three ways:

• Directly connected — networks attached to the router’s own interfaces.
• Static routes — entered by an administrator. Simple and predictable, but they don’t react to failures.
• Dynamic routing protocols — routers tell each other what they can reach and update automatically when links change. OSPF and IS-IS are common inside an organization. BGP connects the tens of thousands of independent networks — autonomous systems — that make up the internet.

OSPF and IS-IS are link-state protocols: every router floods information about its own links to all the others, so they all share the same map, and each one calculates the cheapest paths from its own position with Dijkstra’s algorithm. Cut some links and watch the routes adapt.

Because BGP decides how traffic flows between networks, a single mistaken announcement can send traffic the wrong way or make large services unreachable. That is why internet-scale outages are so often traced back to routing changes.

Routing gets packets to the right machine. In the final lecture of this module, we look at how TCP and UDP get the data to the right program — and what happens when packets go missing.