Dual-Homed Customer BGP Local Preference Usage

ok, so i'm studying for the bgp exam and i have a question about local preference and how it is used in a Dual-Homed Customer to ISP configuration. Say a customer has two routers peering EBGP to a single ISP. One link is primary (T1) and one is secondary link (fractional T1). My question is would it be most efficient to set up a route-map inbound on the router with the T1 primary connection and match the route-map (via acl) with the default route that the ISP is sending me via EBGP? Then set the local preference higher on the T1 EBGP link so that all traffic within my AS will leave my LAN via the T1 link and not the fractional T1 link. And I know of course that i have to have an IBGP connection on both of my External routers which are peering to the ISP.

I think it is but i thought i would post on here just to make sure my logic is correct. Thanks for your help in advance!

Find more posts tagged with

Comments

keenon

CCIEWANNABE wrote: »

ok, so i'm studying for the bgp exam and i have a question about local preference and how it is used in a Dual-Homed Customer to ISP configuration. Say a customer has two routers peering EBGP to a single ISP. One link is primary (T1) and one is secondary link (fractional T1). My question is would it be most efficient to set up a route-map inbound on the router with the T1 primary connection and match the route-map (via acl) with the default route that the ISP is sending me via EBGP? Then set the local preference higher on the T1 EBGP link so that all traffic within my AS will leave my LAN via the T1 link and not the fractional T1 link. And I know of course that i have to have an IBGP connection on both of my External routers which are peering to the ISP.

I think it is but i thought i would post on here just to make sure my logic is correct. Thanks for your help in advance!

I have done this without the use of Ibgp.

yes you have it correct using an inbound route-map with a match all acl ( if both links are connected to the same router) for the link with the primary neighbor. regardless if these connections are not on the same router the fractional link needs to advertise a med out so its less preferred. i would also use a floating static route to point to the backup in case the primary link goes down.

CCIE in the making

CCIEWANNABE

thanks bro. yeah, i pondered the floating static route idea, but i think that if i have IBGP running and set local pref higher on T1 link and lower on fractional link then if the T1 link goes down IBGP will detect and re-route the traffic through the fractional link with the lower local preference automatically. also i believe the med is used to tell the ISP which link is more preferrable, which would affect inbound traffic to my AS, not outbound traffic like I am trying to manipulate. Man, BGP is one tough subject but very awesome and fun!!

kryolla

I would use AS prepend instead of med to prevent async routing

CCIEWANNABE

yep, i was already working an AS-Prepending lab as well. bgp is sweet. thanks again guys and good luck with your studies!

rossonieri#1

2 routers.

for this purpose, i'll combine both AS-path prepending and a route-map. AS-path prepending is to decay the external route, and the route-map to enforce the internal traffic. of course, a floating static route to the backup link should always be used.

APA

kryolla wrote: »

I would use AS prepend instead of med to prevent async routing

Did you mean asymmetric routing??? Synchronous would mean specific timed routing?

AS prepend can still promote asymmetric routing.... MED and AS-Prepend are both used as influencing attributes for inbound path selection to an AS.

If you're prepending over one link... then the neighbouring AS(ISP) is always going to prefer the shortest AS-Path unless they have tagged weights values or local-pref as inbound policies for your announced prefixes.

kryolla

APA wrote: »

Did you mean asymmetric routing??? Synchronous would mean specific timed routing?

AS prepend can still promote asymmetric routing.... MED and AS-Prepend are both used as influencing attributes for inbound path selection to an AS.

If you're prepending over one link... then the neighbouring AS(ISP) is always going to prefer the shortest AS-Path unless they have tagged weights values or local-pref as inbound policies for your announced prefixes.

yes that is what I meant thanks for catching it. As prepend is the most prefered in AS path selection to influence inbound traffic. Of course you have total control in your AS so you can always use cisco proprietary weight or local pref to override anything else that is why it is number 1 & 2 in best path selection.