Hello Gabriel,
On Mon, Nov 14, 2011 at 04:53:22 -0300, gtolon@inti.gob.ar wrote:
Hi. We are using batman-adv 2011.2.0 with Openwrt Backfire-rc6 on D-Link routers, and we're making some tests with iperf to measure bitrate capabilities between nodes. When we put three nodes aligned we notice that the obtained bitrate between the extreme nodes strongly depends on the batman path between them. To make it clear, we have:
A ---------- B ---------- C
With a dinstance of about 20 meters between A and B, and the same distance between B and C. The problem is that sometimes, A and C get connected directly in terms of batman-adv protocol (checked with batctl o), and when that happens, the bitrates are very poor (less than 1Mbps), like if B wasn't there. In fact we disconnected B and obtained very similar results.
Then we reduced tx power settings on A and C, forcing the B hop between them, and we got much better speeds (~20Mbps). We've read about ELP and think that maybe simple OGM messages are not good to measure link quility between A and C in this example, could that be the problem? In that case is there a way to fix this with actual batman-adv algorithms? Thanks in advance!
Gabriel
I think other people will give you better answer than this one, but just as start: OGM are sent in broadcast, which by definition uses a low rate that implies "better transmission than higher rates". Therefore a link having an high TQ doesn't necessarily has a good quality at "high rates" (as you are experiencing).
In my opinion the problem resides in the fact that batman-adv uses broadcast packets to measure link qualities which leads to the aforementioned problem.
I don't know if ELP would help in this sense because as far as I know it still uses broadcast packets.
Please guys correct me if I am wrong.
Cheers,