BFD with Dark Fibre

[dm2020]

Hi All,

Is there any need to run BFD when using dark fibre between two L3 switches? I have customer that wants to set this up but I would have the thought that the inherent capabilities of the switch to quickly detect a link failure, notify the IGP (OSPF in this case) to tear down the adjacency, and then re-converge as required, will be more than adequate. Will BFD do anything to aid this?

Thanks,

[Mark Malone]

Hi
BFD is much faster than a IGP on its own and can reduce failover to ms much less than the IGP can do by itself , it works with the IGP failover or can be configured independently of it but yes its good to have and a best practice in some scenarios

[Mark Malone]

This doc explains the difference of the failover and how it works basically compared to standalone IGP
https://www.networkworld.com/article/2350201/cisco-subnet/reducing-link-failure-detection-time-with-bfd.html

[Reza Sharifi]

Hi,

Overall BFD is faster than any protocol. Usually, BFD is used with BGP as BGP's convergence is much slower than any IGP (OSPF, ISIS, etc..). It can also be used with IGP to lower the convergence time by 1 or 2 seconds but most of the time you see a big convergence time difference when you apply to BGP.

HTH 

[Rujipars]

I had an experience with some network device (not Cisco) that even the fault is detected by physical link failure, having BFD configured (and it detect the fault by interface event, not keep-alive time out) make the recovery faster.  I think it is better if you would test it out for your own network. 
You may simply run BFD on every link but please beware of the BFD packet rate and number of BFD sessions that the system or a line card can support.

I guess that the path of software triggering by BFD is faster than normal IGP reaction.

What I just said is not about the detection by any protocol message exchange and their frequency between peers, neither BFD nor IGP.