Information Technology Reference
In-Depth Information
BGP Network Core Design Solutions
Three primary options are available for a BGP network core design, all of which provide
the ability to significantly reduce prefix count in the IGP processes with varying degrees of
routing and administrative control:
•
Internal BGP (iBGP) architecture
—This architecture uses a single BGP AS in the
network core. The primary benefit gained by deploying this architecture is the reduc-
tion of routing information in the IGP. The defining characteristic of this architecture
is the complete lack of external BGP sessions, with BGP being used almost exclusively
for prefix transport as opposed to routing policy. This design option provides the least
administrative control for defining routing policy through the core.
•
External BGP (eBGP) architecture
—Each distinct portion of the network has its
own AS. An AS peers only with those autonomous systems to which it is directly
connected. In some cases, there also might be some iBGP sessions in regions with
multiple core routers. The defining characteristic of this architecture is the extensive
eBGP peering that ties together all the regions. This design option provides a more
distinct delineation of administrative control between the regional autonomous
systems.
•
Internal/external BGP architecture
—The network core is its own AS and runs iBGP
on all core routers. The rest of the network is broken into separate autonomous systems
that each attach to the network core to receive transit to other diverse network resources.
The defining characteristic of this architecture is the use of iBGP to build the core, with
eBGP providing connectivity to the core AS via eBGP. This design provides the cleanest
delineation of administrative control between the regional networks.
For each of the three BGP architectures presented, the following criteria are applied to
evaluate the strengths and weaknesses of each:
•
Path selection
—What key BGP attributes will be used to determine path selection?
How can the BGP decision process be manipulated to provide the desired routing
policy?
•
Failure and recovery scenarios
—How will the network react to a link or router
failure? What is involved in network reconvergence?
•
Administrative control
—How can BGP be used to define the scope of administrative
control? How does this affect troubleshooting and network expansion?
•
Routing policy
—How can routing policy be defined in this architecture? How
flexible is this architecture?
Throughout the discussion of BGP core architectures, the path-selection process is dis-
cussed in great detail. Instead of reiterating the router ID assignments, they are provided in
Table 5-1. A standard addressing scheme is used to set the router ID for all devices in the
sample discussion network topologies. The router ID for each device is 172.16.X.1, where
X is the router number. The 172.16.13.0/24 prefix is subnetted to provide addressing for all
network core links.
