[MLB-WIRELESS] regarding eBGP - private AS numbers are relatively limited...

David Ashburner d_ashburner at hotmail.com
Thu Dec 16 13:18:28 EST 2004


>That's something I'm going to read up on myself - although I have a
>feeling it's just connected neighbours that need to be specified.

Any instance of BGP has to be configured to know about each directly 
connected neighbour. This means having a neighbour <ip address> as <as 
number>. Within an AS you have a choice of how you get information between 
the machines running BGP. If the AS is small you can add internal neighbours 
(the as number is the same at each machine). So for even a well connected MW 
node you have to have a neighbour for each physical link you have and if 
your AS is a cluster of nodes then you need an entry for each machine that 
is running BGP.  - probably not to bad really.

On the backbone you would want to run something else and not worry about 
iBGP, a border machine communicates out using BGP and inward using RIP / 
OSPF / whatever.  Most routing daemons can handle multiple protocols ( I 
like bird http://bird.network.cz  by the way, simple and very small )

The interesting thing about BGP is that it opens a TCP connection to each 
defined neighbour, exchanges intial routing data, then the connection 
remains and only changes are propagated. This means you don't get the 
constant "chatter" of a lot of the internal routing protocols. The down side 
is once it has decided a route then there is no re-testing and load 
ballancing. Not really an issue for the types of links that would use it  
for now.



To unsubscribe: send mail to majordomo at wireless.org.au
with "unsubscribe melbwireless" in the body of the message



More information about the Melbwireless mailing list