[Olsr-users] olsrd settling time

Eric Malkowski (spam-protected)
Wed Oct 29 18:43:49 CET 2008


You've got me curious - I've always run my mesh network with each node 
having an IP all on the same subnet with the underlying wireless being 
setup for Ad-hoc mode (Atheros MADWIFI 0.9.4 branch).  I've found that 
using the standard broadcast address for the single subnet works just fine.

i.e. If I were to do your configuration I would do something like:

Node 1:   10.5.0.1  mask 255.255.255.0  broadcast 10.5.0.255  (on 
interface ath1)
Node 2:   10.5.0.2  mask 255.255.255.0  broadcast 10.5.0.255  (on 
interface ath1)
Node 3:   10.5.0.3  mask 255.255.255.0  broadcast 10.5.0.255  (on 
interface ath1)
Node 4:   10.5.0.4  mask 255.255.255.0  broadcast 10.5.0.255  (on 
interface ath1)

I then have separate radios that are 2.4 ghz subnets configured as HNAs 
in olsr configuration.  They use the strategy 10.2.X.1 where X is the 
node number and they are each on interface ath0.  So for Node 1 he would 
use 10.2.1.1 on ath0 for his presence on the 2.4 ghz subnet and hand out 
IP addresses in DHCP range 10.2.1.100 - 10.2.1.200 for anyone who wants 
to get on that node w/ standard 2.4 ghz gear.  I run my mesh at 5 ghz to 
stay out of the way of the 2.4 stuff.  The other nodes would realize 
through HNA that subnet 10.2.1.0/24 is available via 10.5.0.1 on the mesh.

I'm curious why you needed to set your broadcast address to an all 1's 
setup (255.255.255.255) to make things work?

The only situation I could think of is if a bunch of ad-hoc nodes are on 
** different ** subnets but still "peering" with OLSR.
i.e.

Node 1:   10.5.0.1  mask 255.255.255.0  broadcast 255.255.255.255
Node 2:   10.5.0.2  mask 255.255.255.0  broadcast 255.255.255.255
Node 3:   10.5.1.1  mask 255.255.255.0  broadcast 255.255.255.255
Node 4:   10.5.1.2  mask 255.255.255.0  broadcast 255.255.255.255

I could see the need to have a couple of different Mesh subnets to 
logically divide up the network, but if a lot of nodes are needed beyond 
a Class C for instance, a Class B setup could be used.

The above seems odd to me and note that I've never tried it -- so don't 
assume it would work.  I'll be curious to hear what the folks on the 
list think.

-Eric





More information about the Olsr-users mailing list