[OLSR-users] a little feedback and questions

[aaron]

hi Andreas, hi lists!
(decoy: basically this sums up my questions from before)


=== contents ===
1) stupid hw / layer 2 / layer 3 (?) question
2) scalability model of olsr available?

---- snip ----

question 1:

so, in vienna we managed to bring a few nodes(rooftop boxes) to olsr-land. 
if everything goes ok this will be a permanent setup with hopefully many 
new nodes joining. quite some pressure to make this work big over here.
So, we will connect new nodes over the time and you can see a little bit 
of the setup at http://funke.lo-res.org -> click on Topologie
So far it seems to be ok except for two nodes (1 * linksys wrt54g and 1 *
debian and orinoco gold with pci - pcmcia adapter as far as i remember).

the setup is like this:

la17 --> oag18 ---------> vivi --------> fiber backbone
client   olsr + bridge    olsr + router

ignore the names (la17, oag18, ...) just nodes.
la17 is the pc with orinoco, oag18 is linksys, vivi is a linksys

what happens is that regularly whenever la17 sends out some bigger 
requests or receives something bigger than the 64 byte ping packet, the 
connection drops down to a few kB/s. With normal ping packets we would 
have theoretially a thruput of 2-11mbit (signal quality is very good!
channel is set to 2 mbit).

any ideas where this could come from? any knobs where we could turn at 
olsr or do you think we have a layer 2 problem?

question 2:

anyway... en route to debugging this problem we started to discuss the 
scalabilty issue of olsr :)

There seem to be two opinions here: 1) meshnetworking bandwidth will scale
with O(1/n) where n is the number of nodes and 2) meshnetworking bandwidth
will scale better than that :)
question: are there any mathematical models for olsr? would somebody be 
interested in discussing models?

our model assumption here is that we connect meshnetworking boxes (olsr 
boxes) with normal rod-style dipoles and where necessary (to bridge larger 
distances) we use yagis.

best regards,
aaron.