Networking Forum powered by InfoSec Insitute

[Steve]

Comments for Decreasing BGP Failover Time Using IP SLA.

[Vito_Corleone]

This is interesting, but it only addresses one side of the peer relationship, which means things are still slower than they could be. BFD is the answer here. Hopefully more providers will support it in the near future.

[mellowd]

BFD for sure.

Good article though

[kannies]

I had no idea that BFD existed!

http://www.cisco.com/en/US/docs/ios/12_ ... s_bfd.html

But looks like a more elegant solution.

At our work, it doesn't seem to be used on customer PE-CE solutions but it is enabled throughout on our Core through every backbone link.

[mellowd]

IP Sla might be a better solution for certain environments, like peering points. Generally ISPs will not run BFD on their peering point interfaces as they'll have 400+ peers. However 400+ IP Sla operations seems insane as well...

[Steve]

You guys are saying IP SLA won't work because the provider's neighborship would remain up and they'd continue to send your traffic on the downed circuit, right?

And then you're saying that BFD is a great solution to this but is not widely supported by ISPs, right?

So, what is the best solution? Low timers?

[Vito_Corleone]

BFD is the best solution. It's just not widely available. Low timers can be good, but they can also break other redundancy feature, like NSF.

Increasing the time is more fun :

Code: Select all


class-map match-all BGP
 match access-group name BGP
   
!         
policy-map BGP
 class BGP
   drop   
!         
ip access-list extended BGP
 permit tcp any any eq bgp rst
 permit tcp any any  rst
 permit tcp any any  fin
!
control-plane
 service-policy input BGP
!


i bet there are easier ways to do this

Hi this is a nice article. He have a similar set-up except that we are multi-homed to two different ISP with different ASN. Is the cuurent configuration applicable with our set-up? Thanks.