[j-nsp] [c-nsp] BGP route flap damping

[Ang Kah Yik]

Hi all on list,

Thanks to all who have replied.
I have taken a look at the RIPE documents and some of the presentations from
the NANOG archive.

Admittedly, I'm not well acquainted with the discrete mathematics used in
some of the recommend reading materials.

However, considering the inputs from the replies on-list as well as the
materials I have looked through, I think the general discussion about route
flap damping can be summarized as the following:

- Vendor implementations tend to dampen entire prefixes instead of the
specific paths that are flapping
- Vendor default thresholds/penalties are far too strict
- Disabling RFD is probably better for routing stability and troubleshooting
than enabling the current vendor implementations of RFD.
- Disabling RFD should be highly considered if router CPUs can handle it

If there are any more points to add to this list, please feel free to do so.

Otherwise, thanks once more to all on list who have shared their views.




On Thu, Oct 9, 2008 at 4:38 PM, Daniel Lete <daniel.lete at heanet.ie> wrote:

> Hello Ang,
> For what is worth and adding to Ajeet pointers, this is the RIPE
> recommendation on the subject.
>
> http://www.ripe.net/ripe/docs/ripe-378.html#recommendation
>
> Regards,
> Daniel
>
>
>
>
>
>
> "Ang Kah Yik" wrote the following on 09/10/2008 03:32:
>
>  Hi Ajeet,
>> Thank you for your reply.
>> Yes, we are multi-homed to our main upstream AS, as well another upstream
>> provider.
>>
>> I have taken a brief look at the paper you recommended but have not yet
>> had
>> the time to digest the information in it.
>>
>> Meanwhile, can we assume (in general) that the conclusion to my original
>> post is that route flap damping is more of a "legacy feature" these days
>> and
>> we can, to a larger extent, disregard it?
>>
>> On Tue, Oct 7, 2008 at 11:56 PM, Ajeet Bagga <bagga_ajeet at emc.com> wrote:
>>
>>  On Oct 7, 2008, at 7:40 AM, Ang Kah Yik wrote:
>>>
>>>  Hi,
>>>
>>>> Thanks for sharing your opinion on the disabling of damping as a BCP.
>>>> Yes, this is something that we've taken into consideration.
>>>>
>>>> However, route flap damping is still in use in a number of networks out
>>>> there.
>>>> Thus, we would like to obtain feedback on how the damping of a flap by
>>>> a transit provider may affect our connectivity.
>>>>
>>>>  Are you multihomed to this transit? To other upstreams? Depending on
>>> the
>>> RFD implementation, withdrawal triggered suppression will indeed affect
>>> your
>>> connectivity. For analysis of arguments against RFD, specifically how it
>>> applies to your case, read the sigcomm presentation, <
>>> http://conferences.sigcomm.org/sigcomm/2002/papers/routedampening.html>.
>>> White paper is available via the ACM portal, <
>>> http://portal.acm.org/citation.cfm?id=633047>.
>>>
>>> ~
>>> Ajeet Bagga
>>> Sr. Network Engineer
>>> Cloud Computing Infrastructure and Services
>>> EMC
>>>
>>>
>>
>>
>>
> --
> Don't Forget to register for our 8th Annual Networking Conference
> 13th and 14th November 2008: http://www.heanet.ie/conferences/2008/
> -
> Daniel Lete Murugarren
> HEAnet Limited, Ireland's Education and Research Network
> 1st Floor, 5 George's Dock, IFSC, Dublin 1
> Registered in Ireland, no 275301  tel: +353-1-660 9040  fax: +353-1-660
> 3666
> web: http://www.heanet.ie/
>



-- 
Ang Kah Yik (bangky) - http://blog.bangky.net