Dell S4048T, Руководство по настройке

Страница: 208 / 1277

You can group autonomous systems into three categories (multihomed, stub, and transit), defined by their
connections and operation.
• multihomed AS — is one that maintains connections to more than one other AS. This group allows the
AS to remain connected to the Internet in the event of a complete failure of one of their connections.
However, this type of AS does not allow traffic from one AS to pass through on its way to another AS. A
simple example of this group is seen in the following illustration.
• stub AS — is one that is connected to only one other AS.
• transit AS — is one that provides connections through itself to separate networks. For example, in the
following illustration, Router 1 can use Router 2 (the transit AS) to connect to Router 4. Internet service
providers (ISPs) are always transit ASs, because they provide connections from one network to another.
The ISP is considered to be “selling transit service” to the customer network, so thus the term Transit AS.
When BGP operates inside an AS (AS1 or AS2, as seen in the following illustration), it is referred to as Internal
BGP (IBGP Internal Border Gateway Protocol). When BGP operates between ASs (AS1 and AS2), it is called
External BGP (EBGP External Border Gateway Protocol). IBGP provides routers inside the AS with the
knowledge to reach routers external to the AS. EBGP routers exchange information with other EBGP routers
as well as IBGP routers to maintain connectivity and accessibility.
Figure 22. Internal BGP
BGP version 4 (BGPv4) supports classless interdomain routing and aggregate routes and AS paths. BGP is a
path vector protocol — a computer network in which BGP maintains the path that updated information takes
as it diffuses through the network. Updates traveling through the network and returning to the same node are
easily detected and discarded.
BGP does not use a traditional interior gateway protocol (IGP) matrix, but makes routing decisions based on
path, network policies, and/or rulesets. Unlike most protocols, BGP uses TCP as its transport protocol.
Since each BGP router talking to another router is a session, a BGP network needs to be in “full mesh.” This is
a topology that has every router directly connected to every other router. Each BGP router within an AS must
have iBGP sessions with all other BGP routers in the AS. For example, a BGP network within an AS needs to be
Border Gateway Protocol IPv4 (BGPv4) 208