[c-nsp] Dual-homing scenario

Brian Turnbow b.turnbow at twt.it
Thu Feb 22 09:00:57 EST 2007


Hi Vincent
Why not set up a  feed from both and filter routes instead of a feed from one and only a default from the other?
That way you can use bgp to pick your paths on both links. 
You can do it 2 ways by as path, which will give you the "closest" matches to your upstreams, or by prefix length ,which will give
You the largest networks. 

Basically you can ask both to send you full table, and then accept what you want.
It will use more memory for BGP , but less for ip route/cef tables .

1 accept based on prefix list (for exapmle: ip prefix-list ONLY20 seq 50 permit 0.0.0.0/0 le 20 , would accept only /20 and below)
1/2 the prefixes now on the internet are /24s and above so if you filter even by /23 on the prefix list you will half the routing table.

or as path list (for example: "ip as-path access-list 1 permit ^1_[0-9]*$"  would permit as1 + all directly attached ases)




Then use your default route for the rest.



Brian





-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Vincent De Keyzer
Sent: giovedì 22 febbraio 2007 13.37
To: cisco-nsp at puck.nether.net
Subject: [c-nsp] Dual-homing scenario

Hello,

 

let a dual-home BGP network. ISP1 comes in with a T3 and sends a full BGP
table; ISP2 comes with a FE and a default-route (for memory reasons).

 

ISP2 is preferrable because of capacity, but ISP1 is preferred because any
route is more specific than a default route; hence all upstream traffic goes
to ISP1, which is not good.

 

Still, it would be nice to keep sending *some* traffic to ISP1; but most of
the traffic should go to ISP2.

 

One way I see to achieve this would be to local-pref based on AS path: if
the route comes from ISP1's AS, or from an AS directly connected to ISP1,
local-pref it with 130; else local-pref it with 100 (and local-pref the
default route from ISP2 with 120).

 

Is this a good way to do it? 

 

Take into account that ISP1 is not very flexible (and ISP2 is flexible, but
likes simple things). :-)

 

Vincent

_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/



More information about the cisco-nsp mailing list