Hi all,

I had some strange results in my testing:
Scenario: 2 nodes, with normal configuration, Ubuntu 12.4, kernel 3.x, HW: usb 
DWL-G122, wCompat 3.5, less noisy channel of all at the room, static nodes. 
Room with no other radios.

- Trasfers from node_a to node_b: 
"iperf -c node_b -u -b5M -t 60" achieve high lose rate med (33%), max BW: 3 Mbps
"iperf -c node_b -u -b10M -t 60" achieve high lose rate med (25%), max BW: 6  
Mbps
"iperf -c node_b -u -b15M - t 60" achieve high lose rate med (21%), max BW: 8 
Mbps

if I check with "-i 10" I can see how the rate in each is going up and down 
(40% of range)

- If I check "iw mesh0 station dump" I can see how the rate is changing around 
all the values from 54 to 5.5 , to 1..., i mean changing very fast for this 
enviroment, is like the negociation of nodes is not stable. During transfer the 
values are almost stable.  Is it a nomal issue?

- Now I fix the bitrate in all the nodes: iw dev mesh0 set bitrates legacy-2.4 
X" where X is 6, 9, 18 ..54
With 1 Mbps I have iperf transfers of 0.6 Mbps without loses, and transfers of 
1Mbps with 26% of loses.
With 6 Mbps I have iperf transfers of 4 Mbps without loses, and transfers of 
5Mbps with 7% of loses.
With 18 Mbps I have iperf transfers of 15 Mbps, and here the TX limits the 
speed of trasmission to 8.8Mbps without loses. 
With 24 is similar to 18, I get a limitation in transmission (like b band to 
11Mbps).RX is capable of reception all without loses.
¿Is any way to disable this limitation without HT modes?

- So if I try in automatic rate mode to transfer 15Mbps I get a 21% of loses 
and 8Mbps of rate  and if I select fix rate I get no loses and 9.3 Mbps of bit 
rate.... Is this normal? 

- If I fix to 6 Mbps the bit rate I can get 4.4 Mbps without loses, but if I 
fix to 5.5 I get 4Mbps with 35% of loses. 
If I select 12 Mbps all is working properly, but if I select 11 I achieve 30% 
of loses.
Reading this, Could be possible that in any part of the code the system is 
using incorrectly the 5.5 instead of 6 or the 11 instead of 12 and it makes the 
lose rate to grow? Any other idea?

cheers,
David
_______________________________________________
Devel mailing list
[email protected]
http://lists.open80211s.org/cgi-bin/mailman/listinfo/devel

Reply via email to