[OLSR-users] Benjamin's negative weight patch/Asymmetric routes

[Benjamin Henrion]

Travis Mikalson <(spam-protected)> [060530]:
> Hi, I just subscribed to this list. I have been experimenting with OLSR 
> for the last few days and I intend to start using it in production shortly.
> 
> I saw Benjamin Henrion posted a patch on March 29th for negative weights 
> which would theoretically enable us to create asymmetric routes on 
> purpose if I'm understanding him correctly.
> 
> The URL he provided doesn't work, though.
> http://www.ffii.org/~zoobab/bh.udev.org/filez/projects/olsr2chan/

Sorry, my account is on hold because I had put too much content on it.

All the patches are here also:

http://www.reseaucitoyen.be/wiki/index.php/Olsr2Interfaces

> Anyway, I am interested in forcing OLSR to use asymmetric routes on 
> purpose to create a "full-duplex" sort of connection. Two wireless links 

Me too :-)

> between two sites, each link handling the traffic in only one direction. 
> Of course I'd like to use OLSR and weights rather than static routes so 
> if one of the two wireless links goes down it will start passing traffic 
> in both directions over the remaining link.
> 
> Site1SystemA--[wireless]-->Site2SystemA
> |                                     |
> |[ethernet]                           |[ethernet]
> |                                     |
> Site1SystemB<--[wireless]--Site2SystemB
> 
> I use four separate systems above for some nice redundancy. A simpler 
> demonstration using two systems instead of four:
> 
>        /wpci0|--[wireless]-->|-wpci0\
> SystemA                              SystemB
>        \wpci1|<--[wireless]--|-wpci1/
> 
> Any ideas for accomplishing this?
> I experimented with hysteresis/weight and LQ/LinkQualityMult, the route 
> always remains symmetric.
> 
> Of course when I go into production I like LQ and intend to use it 
> instead of hysteresis, but if a negative weight via Benjamin's patch is 
> the only way to accomplish asymmetric routing I would be stuck not being 
> able to use LQ anywhere on my network.

My patch does not work in fact, because you probably need to replace
Dijskra algo by BellmanFord, which handles negative weights on graphs.

If there are people who can patch OLSR to support BellmanFord (see also:
http://en.wikipedia.org/wiki/Bellman-Ford_algorithm), I would be able to
sponsor it a bit with some EUR.

--
Benjamin Henrion <(spam-protected)>
http://bh.udev.org
<<<                      INSPIRE Directive will close                >>>
<<<            Public access to State-collected Geographic Data      >>>
<<<                            All over Europe                       >>>
<<<                        http://publicgeodata.org                  >>>