Routing is how a packet finds its way across many networks to a destination it has never been to. Each router knows only its neighbours and a table of prefixes, yet together they build paths across the whole internet, and reroute when a link fails.
Key concepts
- Forwarding versus routing: forwarding is the per-packet lookup in a table; routing is the process that fills the table.
- Longest-prefix match: the most specific matching route wins, and a default route catches the rest.
- Interior protocols: OSPF and IS-IS run link-state algorithms, where every router learns the map and computes shortest paths.
- Distance-vector: routers tell neighbours what they can reach and at what cost, as RIP does.
- BGP: the path-vector protocol between organisations' networks, where policy matters more than shortest path.
Vocabulary
- Autonomous system: a network under one administrative control, identified by a number.
- Next hop: the neighbour a packet is handed to on its way.
- Convergence: the time routers need to agree again after a change.
- Metric: the cost a protocol minimises, such as hops or link speed.
- Static route: an entry written by hand instead of learned.
What to study, in order
- Read a routing table on your own machine and predict where several destinations go.
- Dijkstra's algorithm and Bellman-Ford, implemented on a small graph.
- Build a lab of virtual routers, with namespaces or an emulator, using static routes first.
- Run OSPF between them, pull a link down and watch convergence.
- Learn BGP's basics: sessions, prefixes, path attributes and why a bad announcement spreads.