On 19 Feb 2014, at 18:32, Dino Farinacci <[email protected]> wrote:

>> Well it is not just theoretical, it can happen as soon as you have two
>> egress points, even in active/active mode.
>> 
>> Imagine you have two identical egress points in your network (let say
>> that they can both reach the whole Internet) depending on the IGP the
>> traffic from a part of the network will go to one and the traffic from
>> the other part of the network will go to the second router.
>> Unfortunately the set of destination is not the same in both part of
>> the network so when you fallback to one router after the outage of the
>> other one, you will have misses.
>> 
>> We have evaluated that on our network (that is primary/backup) and
>> simulated it as active/active and noticed that the storm would not
>> negligible.
> 
> Yes, understand Damien. One has to decide how much machinery one puts in a 
> protocol for events that don't happen often.
> 

I could offer you a solution to avoid packet drops, but you might not
like.  It starts with a "b" and finishes with a "uffer" :-)

> Dino
> 

_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp

Reply via email to