[Olsr-users] Calculating LQ in 0.5.0 (Was: olsrd-0.5.0: LQ largerthan 1.0)

[Sven-Ola Tuecke]

Hi Aaron,

before voting "bullshit" based on paper + pen, you should run that thing on
a real node. I'm more the practical guy as you know - and I'm not out for 
academical merits. We all use that little change, since I included this 
within all packages / firmwares I've published since mid-2006. As usual - 
there's plenty of room for mistakes of course.

Should be easy to compile two different versions of olsrd. Run for 10 
minutes. Then estimate, if the ETX value reaches faster the expected 
long-time-ETX-value or not. I remember also having trouble to do the math 
and fiddled with a "fast-start effect" - wild ETX changes in the first 
minute(s) - the reason for the factors.

// Sven-Ola

"Aaron Kaplan" <(spam-protected)> schrieb im Newsbeitrag 
news:(spam-protected)
>
> On Jun 1, 2007, at 11:19 AM, Joerg Pommnitz wrote:
>
>> Hello all,
>> can somebody explain the thinking behind the new formula for
>> calculating the link quality? In 0.4.10 it is
>> calculated this way:
>>
>>   entry->loss_link_quality =
>>     (float)(entry->total_packets - entry->lost_packets) /
>>     (float)(entry->loss_window_size);
>>
>> while in 0.5.0 the formula changed to:
>>
>>   entry->loss_link_quality =
>>     (float)(entry->total_packets - entry->lost_packets) /
>>     (float)(entry->loss_window_size < (2 * 4) ? entry-
>> >loss_window_size:
>>     4 * ((entry->loss_window_size / 4 - 1) * entry->total_packets +
>> entry->loss_window_size) / entry->loss_window_size);
>
>
> IMHO this formula is bogus. I did some calculations on paper and I
> don't see any benefit.
> Maybe sven-ola can explain more about it?
> Otherwise I would vote to revert that one.
>
> all the best,
> a.
>
>


--------------------------------------------------------------------------------


> -- 
> Olsr-users mailing list
> (spam-protected)
> https://lists.olsr.org/mailman/listinfo/olsr-users
>