[Olsr-dev] reduce OLSR bandwidth by tuning emission intervals

[Teco Boot]

Op 26 feb. 2013, om 22:26 heeft Henning Rogge <(spam-protected)> het volgende geschreven:

> On Tue, Feb 26, 2013 at 10:22 PM, Teco Boot <(spam-protected)> wrote:
>>> Because many MANETs might have more than one VPN endpoint. Which makes
>>> it difficult for the backbone to decide which endpoint to use (same
>>> prefix for multiple endpoints).
>> 
>> My described network design does't run routing on the tunnel. Each router sets up its own tunnel. So for example a MANET with 20 routers and 5 exit links has 20 tunnels. Average number of tunnel keep-alives on exit link is 4. For MANETs with few exit links, this models doesn't work well. But this becomes less common, I think.
> 
> Not sure I get how this works.

Think of OSPF stub area's. Inside the stub, only local topology is available. Backbone knows about all prefixes and forwards to border router for destinations in a stub.
In my model, the VPN server is the border router that connects many stubs to the backbone. A stub is a single router with its prefix. Routers that do not have the VPN are not reachable from or thru the backbone, unless they "borrow" an address form another router (and acts as attached hosts). I have a /24 for each router, there is a route to that prefix on each VPN tunnel.
From router to backbone is just a static, added by the OpenVPN client.

Did this help? Agreed routing protocol overhead is decreased to zero?


> 
>>> Thats just a matter of routing metric. If the VPN costs are low
>>> enough, they will be highly preferred.
>> 
>> No, in the described model there is no routing on the tunnel.
> 
> Okay, maybe "HNA with low metric". So that these "external prefixes
> behind the tunnel" have a well defined cost from the mesh point of
> view.

Two options here: reachability for node self (which I do using certificate) or reachability for all nodes in the MANET cluster. For the latter, only the border routers should set up the tunnel (may be your preference). But in a denser uplinked cluster, many uplinks have the overhead of all these reachability info. If reachability info is send reliable (like BGM, OSPF), overhead would be kept low.
Common: no downlink reachablity.

Teco

> 
> Henning Rogge
> -- 
> We began as wanderers, and we are wanderers still. We have lingured
> long enough on the shores of the cosmic ocean. We are ready at last to
> set sail for the stars - Carl Sagan