[c-nsp] OSPF on Ring Networks

[Nick Hilliard]

On 14/05/2011 16:05, Christian Meutes wrote:
> Here it depends. If the next-hop changes OSPF will in most cases converge
> faster than BGP. So if there aren't that many routes and you can live with
> less policy possibilities I would prefer OSPF over BGP for carrying
> non-infrastructure networks.

yes if the next-hop changes, but network topology change events will often 
not cause a next-hop change, just a SFP recalculation - which means that if 
your SFP domain is smaller, it will converge faster.  I.e. the router's FIB 
may be updated, but ibgp will not be aware of the topology change.

The things that cause next-hop changes are customer edge flaps.

As regards carrying non-infrastructure prefixes in OSPF, it really depends 
on how big your network is.  For small networks it doesn't matter very 
much.  For large networks, there is no choice other than ibgp because 
having a functional prefix management system is more important than shaving 
a couple of seconds off a customer edge link-up event.  For networks in the 
middle, they usually start out by putting non-infrastructural prefixes into 
the IGP but change later on because it doesn't scale well and having a 
flexible interior routing policy becomes more important than fast convergence.

> BFD, MPLS-FRR, SPF- and Flood-Tuning for even faster convergence.

...and for even more CPU load.

Nick