Tue Dec 18 18:17:03 CET 2007
>> AHCP was carefully designed to be independent of the routing
>> protocol: it has its own flooding mechanism that doesn't rely on IP
> How does this flooding mechanism work?
There's a rather detailed specification on
The goal of the protocol is to flood a static block of configuration
data throughout a mesh network. Let E be the ``expiry time'' (20
minutes by default), and D the diameter (in hops) of the network.
Then the protocol makes the following guarantees:
- any stale data will expire from the network after at most E seconds ;
- an isolated node will remain configured for a time equal to E ;
- if the network is connected, fresh data will supersede any stale
data within some time between 10 * D and E, depending on the
density of the network and the amount of packet loss (good
behaviour when either the network is dense or the packet loss low).
Basically, every time the configuration changes, the new data is
flooded unreliably through the network (with no retransmissions), and
reflooded every 1/3 E seconds.
Since flooding of fresh data is unreliable, there is the need for some
mechanism for making sure that stale data expires reliably. I'll let
you check the details, but the expiry mechanism is designed to be
resistant to clock skew and to incorrect clocks (but not both at the
The rest is pretty standard -- when a node joins the network or loses
its configuration due to staleness, it broadcasts a request for fresh
configuration data, resending with exponential backoff. Any neighbour
may reply, with both the sender and the receiver checking for staleness.
More information about the Olsr-users