[OLSR-users] generic traffic shaping to priorize olsr messages

[John Gorkos]

Jens Nachtigall wrote:

>Hi,
>
>at the MITRoofNet they discoverd that their ad-hoc net worked ok, when 
>their was little traffic. With loads of traffic the routes got lost. 
>Why? Because the Routing messages by the ad-hoc routing protocol did 
>not come through anymore. I expect the same to happen with olsr.
>
>Therefore, I thought of writing a traffic shaping script, based on 
>hierachical token bucket (htb) and tc. The shortcoming of this idea is 
>that it will only work on linux (which needs to be compiled with htb 
>support). Any other ideas on how to overcome the above problem (for 
>other OSes) in a generic way?
>
>Jens
>  
>
I would argue with this analysis.  In fact, the very reason that I chose 
OLSR versus AODV for my mesh network was based on this:  AODV is an 
on-demand routing algorithm.  A station that has not put network traffic 
across the mesh "recently" must rediscover the network path to the 
take-out point before it can send new traffic.  If the network is so 
busy that those discovery packets get lost, and the packet can not be routed
  OLSR, on the other hand determines the route in advance of a request 
for a packet to traverse the network.  The routes are determined when 
network traffic allows packets to traverse the network.  If, in a future 
situation, there is too much traffic for the route discovery, the 
existing routes continue to be used.  Of course, those routes will 
expire after a certain period of time, but that can be altered in the 
configuration if you anticipate extremely heavy network usage.

Just my thoughts, Andreas will tell us if I'm right...
John Gorkos