[c-nsp] OSPF on Ring Networks

[Christian Meutes]

On May 15, 2011, at 12:48 PM, Nick Hilliard wrote:

> if that happens, then your loopback will disappear in no more than dead-time seconds, or possibly earlier if your igp notices that the interface carrier drops.  This will cause the ibgp next-hop will be become unreachable or else to point by default to the nearest traffic sink on your network.   Either way, convergence time won't be hugely different if you compare ibgp+loopback igp vs non-infrastructure igp prefix situations, because in both cases, all you're doing is updating your fib with an igp-induced topology change.
> 
> Some time after this, ibgp will notice that the edge device has disappeared, at which point the non-infrastructure prefixes will disappear from ibgp.

I don't think that its the same and just depends on FIB operations, because when the iBGP next-hop disappears, BGP will notice that through BGP next-hop address tracking (can react quite fast on the IGP event) and then needs to run its path-decision process to install alternates. I believe that this takes more time than just running SPF, because its like two operations and not only one. If you have PIC support on your platform available then i would say that it should behave like you mentioned, (mostly). Imho BGP would pre-calculate in such a case a backup-route and the BGP path-decision process wouldn't be needed on next-hop change.

Cheers
Christian