BGP Multihomed - FORCE traffic to a specific IP through router B

[bernardssupport]

I have BGP Multihomed routers. Everything works fine and outbound traffic is being determined via the BGP protocol.  Router A ISP is 50Mbs and Router B ISP is 100Mbs. Traffic goes out both routers as it should.  I have a specific destination IP (say 10.10.190.190) that eats a lot of bandwidth at certain times a day. I want to force all traffic destined for 10.10.190.190 to prefer Router B. As it stands it always chooses Router A (because it has a shorter AS path I'm assuming) but my Router B has more bandwidth.  I set the BGP default local preference on Router B, but that forces all traffic out Router B. How can I force traffic headed to 10.10.190.190 out Router B and continue to let BGP determine the best route for all other traffic? Additionally, all the traffic to 10.10.190.190 comes from a specific IP 10.1.121.21.

[Georg Pauwen]

Hello,

local preference in conjunction with a route map should work:

ip prefix-list OUT_TO_B seq 5 permit 10.10.190.190/32
!
route-map OUT_TO_ISP_B permit 10
match ip address prefix-list OUT_TO_B
set local-preference 250
!
route-map OUT_TO_ISP_B permit 20

Then apply that route map to your neighbor ISP B:

neighbor x.x.x.x route-map OUT_TO_ISP_B in

In the prefix list, make sure the prefix mask matches what you have in your routing table (in my example, I used a /32 mask).

As an alternative, you could obviously just use a static route. Not very scalable though, but if you have just that one destination, you could consider using it:

ip route 10.10.190.190 255.255.255.255 x.x.x.x (next hop of ISP B).

[kingsburyr]

To make the most of your bandwidth, it depends on the kind of traffic.  Are you downloading a lot of traffic from that destination or are you uploading a lot of traffic to that destination?  Depending on which one will determine which method of traffic manipulation would be more effective.  ie local preference for outbound or prepending the AS path so traffic is manipulated inbound from the destination.

Ideally you would configure both in a way to minimize asynchronous routing.  If you modify local pref, your out bound traffic may go out provider A, but if the path is shorter, the return traffic may come back in via provider B.

[paul driver]

Hello

Just like to add --

LP is applicable providing you have IBGP peering if not then the weight path attribute would have a better preference, PBR is another option, However I agree with Kingsburyr regards on the return path, You may be manipulating egress traffic but ingress traffic can still come back asymmetric.

Aspath prepending is applicable here but would be advisable to consult with your ISP prior to doing this.

res
Paul