[Olsr-users] Scaling of olsr with reduced TTL

[Juliusz Chroboczek]

> we all know that olsrd won't scale with, say 10'000 nodes (or more).

Not necessarily -- we're still speaking about a topology table of
a mere few megabytes.  It would need some very careful implementation,
but there's no fundamental reason why it shouldn't work.

I believe that OSPF networks larger than that have been deployed.
Admittedly, OSPF is a different beast from OLSR, in particular because
it uses uses reliable flooding rather than periodic flooding.

> I wonder if one could still have a working network of such size by
> simpling reducing the time to live (TTL) to say 8 or something. So
> each node only knows its neighborhood up to 8 hops away,

You cannot do that in a link-state protocol; link-state protocols
collapse unless all nodes have exactly the same topology information
and route selection policies.

You can do that quite easily with a reachability protocol (distance-
vector or its variants, such as path-vector), but that's a different
story.

                                        Juliusz