What is BGP?

BGP (Border Gateway Protocol) is the inter-domain routing protocol that governs how data packets traverse the public internet. It operates as a path vector protocol, exchanging network layer reachability information (NLRI) between autonomous systems (ASes). BGP version 4, defined in RFC 4271, is the current standard and has been in use since 1994. It is the glue that allows independent networks, such as ISPs, cloud providers, and large enterprises, to advertise their IP address blocks and learn routes to every other network on the internet.

BGP works by establishing TCP sessions on port 179 between routers in different ASes (eBGP) or within the same AS (iBGP). Each BGP speaker maintains a Routing Information Base (RIB) and selects a single best path for each prefix based on a decision process that evaluates attributes like AS_PATH length, MULTI_EXIT_DISC (MED), LOCAL_PREF, and origin type. Updates are incremental and triggered by changes in network topology, making BGP more scalable than distance-vector protocols like RIP. BGP does not have built-in security; the RPKI and BGPsec extensions (RFC 8205) have been developed to address route hijacking and origin validation.

In the broader network stack, BGP sits at the application layer of the TCP/IP model but serves as a routing protocol for the internet layer. It is the only protocol that can handle the scale and policy diversity of the global routing table, which exceeds 950,000 prefixes as of 2024. Edge routers at Internet Exchange Points (IXPs) and ISP peering links run BGP to negotiate which routes to accept or announce, with each AS defining its own routing policies. Without BGP, the internet would fragment into isolated networks unable to route traffic between different administrative domains.