maximum observed performance?

Hi guys.  I've long considered building a batman-adv underlay mesh for
some wISP duties.  The problem I'm running into is that I'm only
seeing ~800Mbps in lab tests on real-hardware, far less in GNS3 tests.

I've got a number of >2Gbps RF links and often multiple links that I'm
currently using ospf ecmp to aggregate.

Anyone have any first hand examples or a published doc showing
multi-gigabit over a layer2 batman-adv?  specifically ethernet
interfaces.  The RF layer is separate layer2 bridged and can handle
iperf of ~2.4Gbps+ (Tachyon and Wave Pro radios).

my current model of ospfv3 works but it's all link state and to get it
back to layer2 I'm spending CPU cycles on bridging and encapsulation.
If I have to spend the CPU cycles, I'd love to switch to batman-adv's
link quality metrics

thanks