[OLSR-users] network problems

[Andrew Hodel]

I am going to try all links at 11M soon, perhaps this will be better. 
I really don't quite understand the 802.11 MAC in the case of ad-hoc
with many radios all at different rates.  It seems to me in auto mode
2 radios would connect with each other at the same link speed, if they
don't you could consider it a bad link.

I do agree, bandwidth based routing would be the way to go, however
how do you measure bandwidth without a large stream being sent and
measured every few seconds?  This would eat up your bandwidth while
trying to figure it out.  Perhaps a look at the iperf code could bring
a few ideas.

Over the past few days I have been going out to the locations having
trouble and installing directional antennas.  Of course this is
helping each link, however it really isn't helping the network as a
whole.

Soon I will setup ptp links on a different band to the tallest and
farthest sites.  Then I will have an HNA node on the network at that
location.  Of course this will work, but there goes the dream of a
true mesh.

Interesting read here: http://www.smallworks.com/archives/00000072.htm



Andrew

On 6/18/05, Holger Mauermann <(spam-protected)> wrote:
> Andrew Hodel wrote:
> > We have tried rate 5.5M fixed to avoid lossy links.  You can't even
> > view 640x480 images on the web, by the time the image is finished the
> > connection has dropped
> 
> Are you sure that 5.5Mbps is good for *all* nodes? Try to set bad links
> to lower rates to make it more stable or set it to higher rates to force
> the use of another route.
> 
> In my mesh all nodes run at different rates. With autorate it was almost
> unusable, even with LQ enabled I have lots of bad links. I tried to find
> the optimal rate for each node, but without some bandwidth based routing
> things are now getting more and more complicated... See my recent posts
> about "LQ Routing and 802.11b Link Speed".
> 
> --
> \-- Holger Mauermann
>  \-- (spam-protected)
>   \-- PGP Key Id: 0x8EA8C301
>