OSPF SPF/LSA Throttling

Unanswered Question
Nov 30th, 2009
User Badges:

Hi,


Is it recommended to configure/fine tune OSPF SPF/LSA throttling in every OSPF running router in ISP backbone network or does it depend on type of devices , like 6500=yes, 12k=no, etc?

I am not able to find any good & convincing  doc on OSPF throttling.

Please help!!


Thanks and Regards

Ashish

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (3 ratings)
Loading.
Giuseppe Larosa Mon, 11/30/2009 - 09:48
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Ashish,

in most scenarios you don't need to tune these timers.

However, if you are challenged to reduce as much as possible the convergence time these timers can play a role.

But it is not so immediate to understand how to change them.


see


http://www.cisco.com/en/US/docs/ios/iproute/command/reference/irp_osp3.html#wp1017249


I don't think it is a question of platform, but a question of what you want to achieve in your network, making a tradeoff between immediate reaction to a change and stability, and it gives you the capacity to differentiate the behaviour when the same LSA changes state multiple times or between consecutive runs of SPF.


Hope to help

Giuseppe

marikakis Mon, 11/30/2009 - 11:19
User Badges:
  • Gold, 750 points or more

I agree with Giuseppe. I am not aware of a rule recommending OSPF timer tuning for ISPs in a specific way or in general (i.e. if you are an ISP tune those timers and tune them like this). You know what they say: if it ain't broke, don't fix it. So, do you experience any specific issue in your network that made you consider tuning those timers? Such as: one core router rebooting and a second one crashing when first router boots or anything like that? Sometimes physical links can experience constant flapping causing e.g. increased CPU utilization in some routers. However, there are trade-offs between stability and speed of convergence as already noted by Giuseppe, plus you have to fix the physical layer problems anyway. Trying to keep your IGP as clean as possible (e.g. avoid redistributing customer prefixes to it and use BGP for those prefixes) is a more understandable procedure than IGP timer tuning, and can help keep your network more stable, since you only have to care about your own backbone links functioning properly when it comes to IGP convergence.

Actions

This Discussion