SMC Networks ES4704BD Manual page 526

When BGP runs in the same AS, it's called IBGP. When in the different AS, it's
called EBGP. Generally, the outer neighbors are connected physically and the inner
neighbors can be in any place of the AS. The difference is finally shown in the dealing
manner of BGP to routing information. The equipments may check the AS numbers of
the Open Message from neighbors to decide treating the neighbor switches as the
external neighbor or as the interior neighbor.
IBGP is used in the AS. It sends message to all the BGP neighbors in the AS. IBGP
exchanges AS routing information in a big organization. Attention, the switches in the AS
needn't be connected physically. Only if the switches are in the same AS, they can be
neighbors each other. Because BGP can't detect route, the route tables of other inner
route protocols (such as static route, direct route, OSPF and RIP) need contain neighbor
IP addresses and these routes are used to exchange information among BGPs. In order
to avoid routing loops, when a BGP speaker receives a route notification from inner
neighbor, it would not notify this route to other inner neighbors.
EBGP is used among the AS, and it transmits routing information to the BGP
neighbors of outer ASes. EBGP need physical connection and share the same medium.
Because EBGP need physical connection, the boundary equipments between two AS are
usually running EBGP. When a BGP speaker receives routing information from outer
neighbors, it notifies these routes to other inner neighbors.
3． Route attribute
BGP-4 can share and query inner IP route table through relevant mechanisms, but it
has its own route table. In the BGP route table, each route has a network number, AS
listing information (also called AS path) that it passed and some routing attributes (such
as origin). The routing attribute that BGP-4 used is very complex, this attribute can be
used as metrics to select path.
4． Route-selecting policy of BGP
When receiving BGP notification about a same route from several neighbors,
selecting the best route need to be take into account after routing filtering. This process is
called BGP route selecting process. BGP route selecting process will start only when the
following conditions are fulfilled:
The switch's route must be next hop reachable. That is in the route table there is
the route that can reach the next hop.
BGP must be synchronized with IGP (unless asynchronism is configured; only
restricted to IBGP)
BGP route selecting process is based on the BGP attribute. When there are several
routes that indicate the same destination, BGP need select the best route to the
destination. The decision-making process is as the following:
526