[Olsr-dev] Thoughts about an ETX-FF-Metric plugin

[Markus Kittenberger]

On 6/19/08, Henning Rogge <(spam-protected)> wrote:

>I discovered that if I calculate the LQ values for each timeslice, I get
bad
>results because some slices don't have enough packages.
>that was the reason for me to think of larger time slots,..

But how is the detection of lost packets, now?
When do you count how much packets as lost, when not receiving any packets?

> The linkquality value is calculated by
>LQ = sum (received[1..7]) / ( sum (received[1..7]) + sum (lost[1..7]) )

this formula wonn`t react fast on dead links,.. )-;
if not counting more than the lost hellos, a communicative link will neeed
nearly full history duration to get an bad ETX if the link dies (gets
totally disconnected)

The old approach gave each time slot the same weight, which was better for
detection of dead links(as long as there are at least some packets per time
slot)
now it`s again based on the number of packets, which i dislike,..

but the history has still fixed duration, which is good (-;

Markus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.olsr.org/pipermail/olsr-dev/attachments/20080619/9a1ffb7f/attachment.html>