Information Technology Reference
In-Depth Information
Flexibility
BGP is a path vector protocol, a form of distance vector protocol that constructs an abstract
graph of autonomous systems for each destination. BGP's flexibility is demonstrated in the
number of path attributes that can be used to define policies. BGP path attributes are
parameters that describe characteristics of a BGP prefix. Because the attributes are what
make BGP unique, they are discussed throughout this topic.
You can define two types of policies for BGP: routing and administrative. These policies
often overlap in their functionality.
You can define a BGP routing policy for either the inbound or outbound direction to affect
route or path selection. For example, an inbound filtering policy can be defined to accept
routes that originate only from the immediate upstream provider and customers of that
provider. With proper setting of some attributes, one path can be made to be preferred over
others. Detailed examples of setting routing policies are provided in the rest of this topic.
A BGP administrative policy defines administrative controls for routes coming into the AS
or leaving the AS. For example, an AS might intend to protect its border routers by limiting
the maximum number of prefixes it allows itself to receive. On the outbound side, as
another example, a border router of a multihomed AS might choose to set its attribute in
such a way that only locally originated routes are advertised.
To enforce policies, BGP uses a three-step process:
Input policy engine
1
Path selection
2
Output policy engine
3
Figure 1-1 illustrates this process.
Figure 1-1
BGP Policy Engines and Path Selection: A Conceptual Model
Updates
from a Pee
r
Adj-RIB-In
Path Selection
Input Policy Engine
Loc-RIB
Output Policy Engine
IP-RIB
Updates to
a Peer
Adj-RIB-Out
