Overview
BGPv4 (RFC 4271) is an interdomain path-vector routing protocol that runs over TCP on port 179. It provides a reliable loop-free routing between different campus domains or Autonomous Systems (ASs). BGP routers (called speakers) form peerings (also known as neighbor relationships) with other BGP peers. These peerings can either be internal (iBGP: within the same AS) or external, (eBGP: connecting two different ASs).
Controls route propagation and the selection of optimal routes, rather than route discovery and calculation, which makes BGP different from interior gateway protocols such as OSPF and RIP.
Uses TCP to enhance reliability.
Supports CIDR.
Reduces bandwidth consumption by advertising only incremental updates, which allows advertising large amounts of routing information on the Internet.
Eliminates routing loops by adding AS path information to BGP routes.
Provides policies to implement flexible route filtering and selection.
Provides scalability.
BGP views an autonomous system (AS) as a collection of routes under the control of a single organization, using one or more IGPs to route packets within the AS. The IGP could also be an IBGP within the AS and could use BGP as the only routing protocol.