Can any body let me know the how BFD can be useful to me for faster convergance at control plane during failure compare to OSPF fast hello feature / HSRP hello time in msec....
Also 6500 which IOS version support BFD for HSRP and OSPF both.
Chintan, BFD can help you declare a neighbor down before a line protocol actually goes down. For this BFD support is expected from the higher level protocols.
This helps in faster convergence, as the higher protocol doesnt really have to wait for the line protocol to go down or the dead interval to expire to reconverge. With OSPF Fast Hellos the dead interval can not be changed to less than 1 sec. which still limits you reconvergence to seconds and not sub-seconds.
HSRP with BFD support is good for the edge LAN deployment as where HSRP is deployed without BFD today as well. It although reduces the CPU load when compared to using the multicast HSRP liveness messages conventionally.
For a 6500 the OSPF fast hello support is there from 12.2SXE with Sup720.
HSRP with BFD support is not there as of now in 6500.
Thanks for the info. Howerver, I'm going ot have 6509 running with sup2 /12.2(18)SXF8. which will have diff neighbour running OSPF and some of BGP.
First question, Will it support BFD for OSPF,BGP ? and does BFD frequent hello in ms will not impact system performance and flooding on link ?
Further, Is it possible to run BFD on interface , if my neighbour doesn't support BFD ?
I'm just try to understand that if i put BFD should not be having any adverse impact.
Thanks for your time.
Chintan, with your current hw/sw combination BFD for BGP support is not there. Although interface/OSPF/ISIS BFD support is there.
BFD hellos will not impact the system performance as far as i have seen in real life on a sup720. And also as per documents it shouldnt consume more than 2% of your CPU cycles.
If you neighbor doesnt support BFD you cannot implement it, as its a keepalive mechanism and has to be 2 way.
Though you need to consider the link speeds and delays before implementing BFD with aggressive timers. Very aggressive (min possible) timers are good for high speed core links.
Implement BFD only if you see a need for quick convergence when compared to your current traditional convergence times in your network.
Thanks for the info. It looks like sup720 will give me good result.
By the way , I heard that FHRP is the HSRP BFD implementation. Is that correct ?
I'm not much aware of FHRP.
Chintan thats correct, FHRP could be any first hop redundancy protocol (for eg VRRP/HSRP). Although for feature lookup HSRP with BFD is documented as FHRP - HSRP BFD Peering.
Thanks for clarification.
I see that HSRP BFD peering supported from 12.4(T) series IOS.
I'm not sure 6500 + sup720 + 12.2(33)SXH (Cisco latest IOS) does have HSRP BFD support or not. As of now it looks like 6500 doesn't have support for HSRP BFD.
Is Cisco going to support HSRP in very near future for 6500 ? Any idea ?
I'm again back to you with some more questions.
I've two 6500 connected(Active/Standby) to IP Box (not a routerbut having routing functinality) via through L2 switch. I do have static routes on 6500. So questions are
1. How to use BFD with static routes in 6500 ?
2. Will BFD detect failure from Active 6500 to IP box(BFD capable) due to internal to L2 cloud and remove static route frim Routing table ?
3. What RFC/Standard for BFD and Cisco following which standard ?
Sorry again for bothering you but would appriciate if you can give your ideas on above qurey.. Thank you.
Chintan, what you are trying to achieve can be done, but only using IOS XR. As currently afaik in Cisco IOS XR supports static router based BFD. IOS doesnt have this support yet.
Probably you can change the 65k to a 12k and load it with 3.4.1 to complete your tests with the external IP tester box you have there.
BFD is still in draft stage, Cisco is compliant with versions of this draft. The latest draft is http://tools.ietf.org/html/draft-ietf-bfd-base-06.
After long time I get back again on this topic. I see that now Cisco support BFD version 1 on 7600 SRB, IOS-XR 3.6.2 and that with echo mode.
What is your suggestion ? Which mode i should use i have in between one layer 2 switch ( PE --l2 ---P)