[c-nsp] OSPF VS ISIS

[Kristian Larsson]

On Mon, Feb 27, 2006 at 02:45:59PM +0100, Vincent De Keyzer wrote:
> 
> > But in a properly designed network, both protocols
> > can converge in sub-second..
> 
> My OSPF network is far from converging sub-second (so according to you it is
> far from being properly designed - fine, I can take criticism :)
> 
> First of all, link failure on Ethernet links takes 30 to 40 seconds to
> detect... is there anyway to speed this up?
As soon as you pull the cable your router will
notice this, we're talking ms.
This will trigger an OSPF calculation which
depending on ospf database sizes takes from a few
ms up to a few tenths of a second.

What you are talking about are when you cannot
detect link failure, ie there's a medi converter
some atm or other equipment in between which
prevent you from detecting the cable braking or
similar.
Then you will have to wait for the hold-time which
defaults to 40 seconds or so depending on
platform. This time ( and the keepalive with it)
can be decreased to sub second, ie 300ms keep
alive and 900ms hold-time which would result in
sub-second failure. It would also put quite som
strain on the CPU, for this there is BFD which
offloads this to hardware.
> 
> Then, even after failure detection, I have the feeling it still takes 3 to 5
> seconds to converge...
> 
> What am I doing wrong?
> 
> Vincent
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/