Best Way to Tune OSPF

[TheDukeofBaghdad]

Hi Friends,

So i have simple question, i need to tune OSPF to provide less than a 50ms convergence in case of fiber failure in the network.

My network is very simple it's only 1 area (area 0 of course) and consist of 3 collapsed core switches (Cisco Catalyst 4507) in full mesh and connected to the access layer switches (Cisco Industrial 3010) via layer 2 Port-Channel (2 links).

I have google it and found a LOT of methods .. so i am here to find the recommended method in my scenario.

Thanks

[Reza Sharifi]

Hi,

How long does the convergence take now before any tuning?

[TheDukeofBaghdad]

Hi Reza,

I don't have control over the network at this moment so i don't really know.

I need to create a draft configuration for the switches and make a trip to the field to update the configuration on the switches.

Thanks

[Reza Sharifi]

Hi,

If the configuration is correct and the switches are in a full mesh, the convergence time usually is vert fast (less than 10 seconds) without any tuning.

HTH

[TheDukeofBaghdad]

Hi Reza,

Thanks for your input, however i disagree with you 10 seconds is no longer consider a "very fast" failover and a millisecond failover time is required in most designs for new networks.

Thanks

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Somewhere on Cisco 's main site, they have a whitepaper (search for OSPF and high availability [?]) that explains how to tune OSPF (and EIGRP) for very fast convergence (with OSPF, there's three timers values you'll want to adjust).

Take careful note how the back-off timers work, as their purpose is to keep things like an OSPF flap from driving your OSPF topology into OSPF meltdown.

The same paper might also mention, you might need to adjust interface bounce timers (which can be important when trying to achieve 50ms).

50ms also assumes you can rely on hardware to notify OSPF of a lost neighbor.  If you need to rely on OSPF timers, even with OSPF fast hello or BFD support, it can be difficult to get under 250ms.

[TheDukeofBaghdad]

Hi Joseph,

Thanks for your respond.

So i believe this command will address the interface bounce timers:

SW(config-if)#carrier-delay msec 0

I will use a point-to-point ospf network type to skip DR/BDR election:
SW(config-if)#ip ospf network point-to-point

Sadly i don't think the 4507 support BFD (bi-directional forwarding detection)

OSPF Support for Fast Hellos
SW(config-if)#ip ospf dead-interval minimal hello-multiplier 5

Thanks

[Joseph W. Doherty]

Unsure about BFD support on any specific model without looking it up, but on a 4507 may depend both on sup model and IOS version.

I believe you're on the correct track for interface bounce timer and OSPF fast hellos.

The other OSPF timers I had in mind are:

timers throttle spf # # #
timers lsa arrival #
timers throttle lsa all # # #

The other OSPF parameter you might want to enable is ISPF.

[TheDukeofBaghdad]

Hi Joseph,

I have added the following commands to the 3 routers i am using in my LAB:

!
router ospf 1
 timers lsa arrival 50
 timers throttle lsa all 10 100 1000
 timers throttle spf 10 100 1000
 timers pacing flood 5
 timers pacing retransmission 60
 ispf
!

Small Test Topology:

R10 & R20 are Cisco 1841 /// R30 are Cisco 1921

R10, Fa0/0 --> R20, Fa0/0

R10, Fa0/1 --> R30, Gi0/0

R20, Fa0/1 --> R30, Gi0/1

Ping Test: I started a ping from R30 to loopback on R10 and disconnected the direct link between the two routers and lost only 1 packet

R30#ping 1.1.1.1 repeat 50000

Success rate is 99 percent (49999/50000), round-trip min/avg/max = 1/1/32 ms

Also i have found that BFD is working on the 4507 but i think it's limited (or at least the documentation said that!) do i still need to use it ? is it going to provide any additional benefits ?

Thanks

[Joseph W. Doherty]

BFD is only really useful when an OSPF link goes down, the OSPF neighbors rely upon OSPF detecting the outage via its keep alive hellos.