What is BGP, and what makes it fundamentally different from IGPs like OSPF? When is BGP used?

Question

Accepted Answer

BGP (Border Gateway Protocol) is the routing protocol that runs the global internet. It is the only Exterior Gateway Protocol (EGP) in widespread use, and it is fundamentally different from Interior Gateway Protocols like OSPF in its design philosophy, scale, and how it makes routing decisions.

The most important distinction is scope. IGPs like OSPF operate within a single autonomous system (AS), which is a network under one administrative domain. BGP operates between autonomous systems, exchanging routing information across organizational boundaries. When your company connects to an ISP, or when two ISPs peer with each other, they use BGP. Every ISP, large enterprise, and content provider on the internet runs BGP.

BGP is a path-vector protocol, meaning it selects routes based on a set of path attributes rather than a simple numeric metric. The most fundamental attribute is the AS-PATH, which lists all the autonomous systems a route has traversed. BGP uses this not only for loop prevention (a router rejects any route containing its own AS number) but also as a rough measure of path length. However, BGP considers many other attributes before falling back to AS-PATH length, giving administrators extensive control over traffic engineering.

BGP uses TCP port 179 for neighbor connections, which is different from OSPF and other IGPs that have their own Layer 3 or 4 protocols. BGP neighbors must be manually configured, there is no automatic neighbor discovery. This manual configuration is intentional because BGP peerings represent business relationships between organizations, and each peering should be deliberately established.

There are two types of BGP peering. eBGP (external BGP) runs between routers in different autonomous systems. These are the peerings that connect your network to ISPs or partners. eBGP peers are typically directly connected, and the TTL is set to 1 by default. iBGP (internal BGP) runs between routers within the same AS. iBGP is used to distribute BGP-learned routes internally. An important iBGP rule is that a route learned from one iBGP peer is never advertised to another iBGP peer (to prevent loops), which means iBGP requires either a full mesh of peerings, route reflectors, or confederations.

BGP is used in several scenarios. Internet connectivity: when an enterprise has multiple ISP connections (multihoming), BGP provides control over inbound and outbound traffic paths. Service providers use BGP to exchange the full internet routing table (currently over 900,000 IPv4 prefixes). Large enterprises use BGP internally to manage traffic between data centers and regions. MPLS VPNs use BGP (specifically MP-BGP) to distribute VPN routing information. Modern data center fabrics often use eBGP as the IGP replacement in a Clos topology design.

BGP's design philosophy of stability over speed makes it converge more slowly than OSPF (seconds to minutes versus subseconds), but this is acceptable for its use case where rapid, uncontrolled convergence could cascade across the internet.