[Olsr-dev] Strategy for ETT metric

[Aaron Kaplan]

On Mar 27, 2008, at 5:03 PM, Hannes Gredler wrote:
> hi henning,
>
> i'd be certainly in favour of #1.
>
> /hannes
>
> On Thu, Mar 27, 2008 at 09:03:52AM +0100, Henning Rogge wrote:
> | Hello,
> |
> | I'm working on a wrapper for olsrd to include layer 2 information  
> into my
> | routing plugin system. I have looked through some papers about  
> ETT and the
> | only sollution to measure link bandwith seems to be the to send  
> small unicast
> | packages along each 1-hop link (unless there is other traffic on  
> this link).
> |
> | I see two options to do this:
> |
> | 1.) wait for some time, then send one package to every 1-hop  
> neighbor
> |
> | 2.) distribute the packages over time
> |
> | Option 1 does wake up the olsrd one once for all neighbors, but  
> it sends a
> | burst of packages that might block some real traffic.
> |
> | What do you think about the two options.

hm... with new proper timer infrastructure - does it make a big  
difference to have both ?
At least right now, we don't know what is better. But I guess #1 is  
better vor mobile devices
that want to sleep a lot and send little and if they need to send,  
wake up once.

(of course, I know this is all a bit hypothetical now).