[Olsr-users] DLEP: reactive routing

Henning Rogge (spam-protected)
Tue Mar 15 08:56:37 CET 2016


On Mon, Mar 14, 2016 at 4:50 PM, s-mtafel <(spam-protected)> wrote:
> Hello,
>
> are you aware of that your dlep_router is not reacting in that it removes
> routes when all destinations to a neighbor are down but waits for the hello
> based timing out of its entries?

Yes, I am aware of this... at the moment DLEP is only used to help
calculating the DAT metric value of a link, nothing more.

> I have this master thesis of kim nilsen (SAAB) in mind that showed DLEP is a
> perfect channel for reactive routing. For very bandwidth constraint environs
> as I am working on, that is quite attractive. Maybe you've disregarded it
> since you are on a less resource constrained environ.
>
> What do you think? Is it down to a extension?

Yes, this could be an extension. Would be quite easy to do, just react
to the change in the layer-2 database (oonf_class allows to add
listeners to the database objects of oonf_layer2) and then modify the
NHDP database.

The potential trouble with "fast link takedown" (e.g. by using a
layer-2 notification) is that you need to send a HELLO (cheap) and a
TC (expensive) when this happens, otherwise you risk a routing loop
because the rest of the mesh will not know that the link is gone.

So if your link goes down because of congestion, you risk to make
everything worse.

Still, might be interesting experiment.

Henning Rogge



More information about the Olsr-users mailing list