BGP's Basic Components
When you are connecting two or more ISPs, network access points
(NAPs), and exchange points
When you are using multi-homing in an enterprise network that con-
nects to more than one ISP
One of the main reasons not to use BGP is if your router can't support the
huge routing tables needed to support BGP. Remember that there are about
7,000 AS numbers in use today, and the router learning those huge tables
requires a lot of RAM and processing power.
In a lab scenario, using two Cisco 2500 series routers works fine since the
tables it has to learn are small. But once the table acquires over 70,000 routes
and gets above 35MB, the router begins to develop serious processing prob-
lems. This increases the latency dramatically to the point where the through-
put through the router is...well...you might as well disconnect the cable to
the router and use sneakernet (placing the data on a disk and walking it to
its destination) because the data will get to its destination faster.
There are some other reasons not to use BGP, including:
When there is a single connection to the Internet. Use a default route
instead. You'll just be wasting bandwidth, memory, and processing
When your network does not have the bandwidth to support the
amount of data needed to be passed, including BGP's huge routing
ngress filtering allows you to decide the routes that you will advertise
to other BGP neighbors or peers. When using BGP in your AS, you have the
ability to announce the routes in your AS that you want to be seen by the
Internet. To safeguard this process, many ISPs have policies in place to
accept the announcements of routes that belong to your AS.
RFC 2267 outlines how ISPs should filter ingress routes and traffic.
Copyright ©2001 SYBEX , Inc., Alameda, CA