Inbound load balancing over 2 T1's

[johnvojtech]

I am sure this has been discussed before, but I cant seem to find it if it has. I have 2 T1's from one carrier terminating into my local 3640. I have somewhat load balancing on outboound traffic leaving my router, but the inbound traffic is always sent on one T1. I saw that if I use the prepend statement should kinda load balance traffic coming into my router. I applied the route map and there has been no changes. All the inbound traffic has been sent over the same T1.

What am I missing? I am not looking for true load balancing, I just want to use both T1's better.

Thanks.

[rais.ahmad]

What happens when you shut off that working T1? Does the other T1 kick in?

I assume you have two different bgp neighbors you are peering with. Your prepending should have affected your incoming trend.

Hope this helps.

[Richard Burts]

The original post says:

I have 2 T1's from one carrier terminating into my local 3640.

So I would not think that there are two BGP neighbors.

Prepending may influence traffic when there are two links to two different BGP neighbors. But if it is two links to the same neighbor then prepending will have no real effect.

Assuming from the original post that he is running BGP on both link I think that his best option is to talk to the provider and ask them about load balancing what they are sending to him. Without their coooperation and direction I think that there is very little he can do from his end to balance incoming traffic.

HTH

Rick

[rais.ahmad]

I think we can still think that there are two BGP neighbor 'routers', though only a single BGP neighbor AS ;-)

Regards.

[johnvojtech]

It is 2 T1's coming from the same ISP (ATT). They both terminate into the router and using BGP, load balance for outbound traffic. If the one T1 goes down, the other takes over fully and no users see any problems. According to the ATT network engineer, I can use prepend or community statements to load balance on the inbound traffic. I used the prepend and it didnt work.

[rais.ahmad]

I would still investigate prepends, perhaps greater AS-length than what you tried before. I don't know how to use community for this purpose. MED seems promising too, but most ISPs remove MED values.

Hope this helps.

[vcjones]

Some things to check...

1 - Make sure you are not sending any community strings with your advertisements that flag the link as a "backup" or "preferred" link.

2 - Make sure that both links are advertising the exact same prefixes. If one is advertising two /24s and the other a summary /23, the /23 will never be used as long as the /24s are available.

3 - If you did not already, try prepending multiple iterations of your prefix on the overused link. If prepending five copies of your prefix does not do the job, something other than AS path length is deciding the link to use.

Good luck and good hunting. If the above three don't show you the path, call ATT again and have their engineer actually look at their router configs.

Vincent C Jones

www.networkingunlimited.com

[johnvojtech]

Thanks for the info. I will try to prepend up to five copies. I tried 3 and saw no luck. I will try that and talk to ATT again and see what the issue is.

Thanks again.

[Richard Burts]

It would help us if we understood a little more about your environment. For example what are you advertising to the provider? Are you advertising a single prefix to ATT representing your network, or if more than one how many prefixes are you advertising?

If you could post the BGP part of the config it would be helpful.

HTH

Rick

[johnvojtech]

router bgp 15303

no synchronization

bgp log-neighbor-changes

network 12.158.15.48 mask 255.255.255.240

network 12.158.24.240 mask 255.255.255.240

network 198.199.191.0

neighbor 12.126.53.97 remote-as 7018

neighbor 12.126.53.97 distribute-list 4 out

neighbor 12.126.53.101 remote-as 7018

neighbor 12.126.53.101 distribute-list 4 out

neighbor 198.199.191.3 remote-as 15303

neighbor 198.199.191.3 next-hop-self

maximum-paths 2

no auto-summary

[olorunloba]

This configuration does not include the AS path prepending. You need to have

router bgp 15303

neighbor 12.126.53.97 route-map SETPATH out

route-map SETPATH

set as-path prepend 15303 15303

[olorunloba]

Rather, reading through the config and the post, I do not think that will work. AT&T router would always select one route, hence by prepending as before, it will only switch the traffic to the other link.

Rather, noticing that you are advertising more than one prefix to your provider I will advice that you prepend those prefixes differently. This will make your provider to have different preference for the prefixes on each T1. Therefore, traffic will flow primarily for a prefix on a link, while the other prefix flow though the other link.

Check the following configuration

router bgp 15303

no synchronization

bgp log-neighbor-changes

network 12.158.15.48 mask 255.255.255.240

network 12.158.24.240 mask 255.255.255.240

network 198.199.191.0

neighbor 12.126.53.97 remote-as 7018

neighbor 12.126.53.97 distribute-list 4 out

neighbor 12.126.53.97 route-map SETPATH1 out

neighbor 12.126.53.101 remote-as 7018

neighbor 12.126.53.101 distribute-list 4 out

neighbor 12.126.53.101 route-map SETPATH2 out

neighbor 198.199.191.3 remote-as 15303

neighbor 198.199.191.3 next-hop-self

maximum-paths 2

no auto-summary

route-map SETPATH1

match ip address 1

set as-path prepend 15303 15303 15303

route-map SETPATH2

match ip address 2

set as-path prepend 15303 15303 15303

access-list 1 permit 12.158.15.48 0.0.0.15

access-list 1 permit 12.158.24.240 0.0.0.15

access-list 2 permit 198.199.191.0 0.0.0.255

I guess this should work.

[johnvojtech]

I had the exact same commands in there, except i had 15303 prepended 3 times. And after a day, there was no inbound traffic on the unused T1.

[olorunloba]

Have you tried the config with different AS-Path for the different prefixes?