[Olsr-users] [Olsr-dev] Question: is possible to use mixed LinkQualityAlgorithm among nodes ?

[Saverio Proto]

Hello,

I wanted to give feedback about this.

We are happy now using etx_float on those nodes where a strong (0.3)
LinkQualityMult is applied. Links do not flap anymore. We are also
using the LinkQualityAging at 0.02.

thanks for support.

Saverio



2012/12/12 Ferry Huberts <(spam-protected)>:
> One note though:
>
> ffeth and ffeth_nl80211 store their values in a different location in the
> 'wireformat' so using that is not compatible with the other lq algorithms.
>
> I have a branch that fixes that for ffeth (but not for ffeth_nl80211 since
> that plugin is experimental), but have not published it (yet) because it's a
> breaking change.
>
> If you want I can publish it though.
>
>
>
>
>
> On 12/12/12 22:52, Henning Rogge wrote:
>>
>> On Wed, Dec 12, 2012 at 7:31 PM, Saverio Proto <(spam-protected)> wrote:
>>>
>>> Short version:
>>> Question: is possible to use mixed LinkQualityAlgorithm among nodes ?
>>> If I have most nodes with "etx_ff" and some other nodes running
>>> "etx_float" I will run into trouble ?
>>
>>
>> No... you only get in trouble if you use "etx_ffeth".
>>
>>> Long version:
>>> In Ninux we have some links that are very long, like 20Km. These links
>>> have very low bandwidth, around 1Mbps, but also very low packet loss,
>>> so that when there is no user traffic the ETX is 1, and all the
>>> traffic prefers these links.
>>>
>>> We use then LinkQualityMult to raise the ETX to 10. Raising so much
>>> causes the link to flap, because with little traffic is very easy to
>>> have a small packet loss and the cost will reach LINK_COST_BROKEN.
>>
>>
>> Try to raise the link quality multiplier to 3 on both sides... that
>> should give you an ETX of 9 without instant link breaking.
>>
>>> I want to try to tune LinkQualityAging that now is set at the default
>>> value to see if I can avoid this flapping, and I want also to try the
>>> etx_float algorithm to see if I can solve this problem.
>>
>>
>> lqaging doesn't work in etxff and etxff_eth.
>>
>> It will not help with links with lqmult 10, because lq<0.1 is
>> considered a "broken" link I think.
>>
>> Henning Rogge
>>
>
> --
> Ferry Huberts