Unlike the realm of IGPs, where religious wars rage between the noble followers of Open Shortest Path First (OSPF) and heretics of Intermediate System to Intermediate System (IS-IS), or vice versa, interdomain routing sits in a realm where only a single protocol rules. Tomes in all forms of media have been devoted to BGP, and there is no shortage of information available to use to develop a plan for merging networks, including BGP, by Iljitsch van Beijnum (O’Reilly), and Internet Routing Architectures, by Sam Halabi (Cisco Press).
Like most of the other topics covered in this book, planning the architecture for BGP in ASs is normally not a greenfield effort. There are few, if any, large networks that do not already have a BGP architecture. In terms of operating, merging, or maintaining a BGP network, planning is key. It is necessary not only to look to the future to see what design is best for scalability, but also to determine which design will merge most easily and scale most efficiently.
According to RFC 1771, all Internal BGP (IBGP) peers must be fully meshed within the AS to prevent routing loops. Using the full mesh has the advantage of ensuring that all routers in the AS have the same topology information. The disadvantage comes with the configuration and scalability of the IBGP peering. To avoid these issues, the following additional BGP architectures have been created:
The division on an AS into multiple sub-ASs with a full mesh ...