[Olsr-users] dampening of OLSR

[Theo Gjaltema]

Hi,

I am using the etx metric.
In the test network we see lots of re-routings (kernel route changes)
with a stationary set of nodes.


Theo.


Henning Rogge schreef:
> Am Dienstag 26 Mai 2009 09:12:58 schrieb Theo Gjaltema:
>   
>> Hi,
>>
>> I am currently investigating OLSR with metrics based on link-costs and
>> see that the  costs of the  routes are varing very fast.
>>     
> What metric are you using ? etx_ff should be a little bit more tame than 
> etx_fpm/etx_float.
>
>   
>> For this reason I am looking for an uniform mechanism of adding
>> hysteresis/dampening to OLSR.
>> In the olsr there is allready a form of hysteresis build-in (checks for
>> a delta of 10%) but this is not flexible enough.
>>     
>
>   
>> My idea is to add a plugin for this removing the fixed hysteresis and
>> adding dampening.
>>     
> You don't need to remove anything, because the "dijkstra hysteresis" is 
> controlled by the routing metric plugin too (there should be a callback called 
> "is....relevant(linkcost1, linkcost2)" ).
>
> Markus Kittenberg played with the idea of a "long time" metric for some time, 
> but we never got enough time to create a good plugin. We would be very happy 
> to see a suggestion for an improved etx_ff. :)
>
> Henning
>