What Happens When a Link Fails

The internet is not one line from sender to receiver. It is a mesh of nodes, routers and the devices that relay data, connected by many overlapping links. Between any two points there are usually several possible paths, not just one. This redundancy is the whole point: it exists so the network has a backup plan built in before anything ever breaks.

When a link fails, say an undersea cable gets cut, the routers next to that break notice that their usual path is gone because expected signals stop arriving. They update their routing information and pass word to their neighbors, who update theirs in turn. Data packets, which already travel independently and get reassembled at the destination, simply get sent down a different sequence of working links. Each packet can take its own route, so rerouting happens packet by packet, on the fly, without any central authority approving the change.

This same design explains why one server crashing does not take down the internet. A server is an endpoint, something hosting a specific website or service. The internet's routing fabric does not depend on that one machine; it only needed it as a destination. So the network keeps running, and only the content that lived on that server becomes unreachable. The limit case is Big Idea 4's next idea: if enough links fail that every path between two regions is gone, no amount of clever rerouting helps. The regions are simply disconnected until a path is restored.