BGP routing protocol, evolution and developments

The BGP protocol, in its version IV, was specified to the IETF in 1995 at the start of the Internet's growth. It is responsible for the external routing of IPv4 unicast routes, and must therefore dynamically determine the best path to reach an IPv4 destination in the Internet (while respecting the administrative routing policies defined by autonomous system operators).

In recent years, IP networks, MPLS and the Internet have evolved considerably:

• in terms of size, from 1995 to 2010 the number of IP routes on the Internet increased 16-fold, reaching 330,000 routes by mid-2010, and the number of autonomous systems (AS) increased 20-fold, reaching 34,000 interconnected autonomous systems;

• in terms of availability, IP networks, which were mainly used for less interactive applications (e.g. email, FTP...) and by researchers, are now used by many highly interactive applications (web, voice over IP, online games, corporate VPNs) and many companies are increasingly dependent on these IP networks (with financial implications in the event of failure);

• In terms of routing, BGP was designed to carry only IPv4 unicast routes. It is now also used for IPv6, MPLS, IP VPN and Ethernet VPN routes.

This article describes the main developments in the BGP protocol over the last ten years, and in particular the BGP extensions that improve scalability, availability, administration and routing management for new address families such as IPv6 and BGP/MPLS VPNs.

Readers will find a summary table of acronyms and abbreviations at the end of the article (see table 5 ).