BGP and the routing structure of the Internet
Though we might not think about it. Any time we are connected to the internet we are relying on packet based routing protocols. There are two main philosophies for routing packets: proactive, where routers actively maintain rough mapping of where to send packets, and reactive, where a route is discovered on the fly when a packet arrives. The former is used for the majority of packet transactions on the internet. ISP’s route packets all over the world using a protocol called BGP.
In BGP, routers maintain a table of where to send a packet next based on the IP it wants to reach. When a new router joins the protocol, it broadcasts an advertisement of is presence to anyone it is physically connected to. It lets them know who is connected to it, possibly linking a previously separate connected component of machines to the giant connected component that is the internet. It can then build tables on where to the path to send packets to next based on the IP address of the destination based on their tables, adding itself as an extra ‘hop’ in the path to the final destination. Around every 30 seconds, connected routers will send a small ‘keep-alive’ message to their physical neighbors. If a router is going down, it will hopefully signal its neighbors, but if it doesn’t, the ‘keep-alive’ messages will allow them to discover the link is down. They will then notify anyone who had gotten a route through that link from their table. It is very important for these routers to have a very low failure rate. They are usually a bridge from the networks they are in front of to the rest of the internet (internet users usually have only one way to connect to the internet, which would go through their ISP).
Sources:
http://docwiki.cisco.com/wiki/Border_Gateway_Protocol
http://www.linux-mag.com/id/1353/