OSPF fail-over convergence

[osw200051]

Hi All,

I have an testing environment for the voice traffic on below diagram.

During the IP Phone A and IP phone B are in streaming, I try to disconnect one of the OSPF link 1 (192.168.1.0/30), the voice traffic would fail-over and use alternate link 2 (192.168.2.0/30).

However, I observed that there are huge amount of voice packet loss (around 100) when the failover occurred.

There are any method to improve the condition?

Thanks

 

[Mark Malone]

Hi

there are methods to speed up OSPF convergence , I would take a quick read of this doc its very well explained with some options you could try to increase your failover speed

http://blog.ine.com/2010/06/02/ospf-fast-convergenc/

[Joseph W. Doherty]

Besides working on OSPF tuning (using something like Mark's excellent reference [Cisco's main web site also has whitepapers on OSPF tuning for subsecond convergence]), you might consider converting the two p2p links, between your two VSS pairs to an Etherchannel link.  Then if one of those link fails, you only have Etherchannel redirection time to deal with, OSPF convergence is a non-factor.

[Rolf Fischer]

Hello,

I can only reemphasize what Joseph has already suggested. The most important VSS design rule in terms of convergence is "dual-attach with MEC" (Multi-Chassis Etherchannel). According to Cisco documentation, hardware-based etherchannel convergence is typically faster than 100 ms!

Also keep in mind that VSS does not synchronize routing protocol specific data-structures like the OSPF LSDB between active and standby switch. The recommendation is to configure non-stop forwarding (NSF) or non-stop routing (NSR) under the OSPF processs.

HTH
Rolf