OSPF Hello/Dead Timers Over WAN Links

[terrygwazdosky]

We have a mix of dark fiber and Metro Ethernet coupled with T1 backup links between our hub and spoke sites.  I'm tinkering with hello timers and am wondering what Cisco recommends as well as what other use on their WAN links.  My plan is to use aggressive timers on the primary links and leave the backup links at defaults.

As a test I setup fast hellos over one of the backup T1s with "ip ospf dead-interval minimal hello-multiplier 5" and OSPF has not dropped once in the 4 days since. 

Any input is welcome.

[Peter Paluch]

Hello Terry,

If your OSPF routers implement BFD then it is generally better to use BFD instead of aggressive timers in OSPF. The reason is that the BFD is more lightweight and requires less processing. On high-end routers like 7600 series and in special line cards, the BFD is even implemented in hardware, thereby offloading the CPU - although this is probably not your case here.

If the BFD is not an option for you then I believe that the two primary things to look for are CPU load incurred by aggressive timers, and the stability of OSPF adjacencies. As a rule of thumb with no technical explanation (just a feeling in my gut), I would be concerned if the CPU load got increased by additional 20% thanks to OSPF Fast Hellos you're utilizing currently. Make sure that the load increase is below that level.

With ip ospf dead-interval minimal you have configured your OSPF router for a dead interval of 1 second. Within these 1 seconds, you send 5 Hello packets. Now you are indicating that the OSPF adjacencies appear to be stable. In that case, you may be just fine with the current settings.

Best regards,

Peter

[terrygwazdosky]

Thanks, Peter.

I should have mentioned that the remote hub sites routers do not support BFD.

Looking at the historical CPU usage in our network monitoring app, I do not see that it has increased at all on either hub or spoke.  I'll try adding fast hellos at few more spoke sites and check the hub again.

[Peter Paluch]

Hi Terry,

That sounds good. There is no absolute "no" related to OSPF Fast Hellos, just the usual concern with CPU usage, bandwidth consumption and the stability of OSPF adjacencies - but if your experience is positive then there is no reason why you should not use it.

Best regards,

Peter