2011/7/11 Max Ip ipmax2011@googlemail.com:
In the two nodes, without using batman-advanced, I got the following :
netperf -t UDP_STREAM -H 192.168.1.3 -c -C -- -m 32768 UDP UNIDIRECTIONAL SEND TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.1.3 (192.168.1.3) port 0 AF_INET : demo
Socket Message Elapsed Messages CPU Service Size Size Time Okay Errors Throughput Util Demand bytes bytes secs # # 10^6bits/sec % SS us/KB
114688 32768 10.00 605 0 15.9 3.35 34.693 114688 10.00 604 15.8 0.70 7.291
So, the received (real throughput as daniele said) is 15.8 between nodes 192.168.1.2 and 1.3. Also, the cpu usage is as expected.
Then, after installing batman-adv 2011.1.0, and naming the new nodes for bat0 interface 10.42.43.1 and 10.42.43.2, I got
netperf -t UDP_STREAM -H 10.42.43.2 -c -C -- -m 32768 UDP UNIDIRECTIONAL SEND TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.42.43.2 (10.42.43.2) port 0 AF_INET : demo
Socket Message Elapsed Messages CPU Service Size Size Time Okay Errors Throughput Util Demand bytes bytes secs # # 10^6bits/sec % SS us/KB
114688 32768 10.00 120574 0 3160.7 51.45 11910.058 114688 10.00 27 0.7 0.95 220.982
Thus the received throughput is 0.7 Mbps. Out of 120574 data only 27 are received. How can using batman-adv decrease the throughput by such a big amount? Is it because of using the accesspoint (managed mode instead of ad-hoc mode)?
I've not clear what happens using batman in this way, with an access point and in managed mode. You can try to examine what is happening using wireshark with another PC or creating a virtual interface in monitor mode, see: http://linuxwireless.org/en/users/Documentation/iw#Adding_interfaces_with_iw You can see if packets pass throught the access point (if it has AP isolation enabled) or go directly and at what rate packet are sent.
Anyhow i really advise you to pace netperf because this can influence the results.
Cheers
MAx
On Mon, Jul 11, 2011 at 1:14 PM, Daniele Furlan daniele.furlan@gmail.com wrote:
2011/7/11 Max Ip ipmax2011@googlemail.com:
Hi all,
I have network connection of two laptops connected through wifi in Access Point mode with bandwidth 54Mbps (802.11g).
Laptop A ---- Access point ------ Laptop B 192.168.1.2 ---- 192.168.1.220 ------ 192.168.1.3
I used netperf to measure the bandwidth and CPU utilization
Without installing batman-adv on Laptop A and Laptop B , the results of netperf are:
netperf -H 192.168.1.3 -t TCP_STREAM -l 10 -c
TCP throughput : 11.89 Mbps TCP CPU usage: 1.73 %
netperf -H 192.168.1.3 -t UDP_STREAM -l 10 -c
UDP throughput: 22.5 Mbps UDP CPU usage: 1.8 %
Now, after I installed batman-adv 2011.1.0 on Laptop A and Laptop B. To the nodes Laptop A and Laptop B I assigned IP address:
sudo ifconfig bat0 10.42.43.1 (laptop A) sudo ifconfig bat0 10.42.43.2 (laptop B)
The protocol runs fine and the originators are seen on both sides. Also, the avahi-discover shows both laptops. When I now tried to measure the throughput,
netperf -H 192.168.1.3 -t TCP_STREAM -l 10 -c
TCP throughput : 9.23 Mbps TCP CPU usage: 3.91 %
netperf -H 192.168.1.3 -t UDP_STREAM -l 10 -c
UDP throughput: 3200 Mbps UDP CPU usage: 51%
The real throughput is what you read in the second line of the results of netperf. The value you have reported refer on the packet sent (first row of resutls) and not to the packets correctly received.
Furtermore it is better to "pace" the load offered by netperf using options -b and -w as you can read in: http://www.netperf.org/svn/netperf2/tags/netperf-2.4.5/doc/netperf.html
Why would the UDP throughput be so high? There is no possibility to have it so high. Is it because of the bad configuration in access point as it is non batman interface or sth else?
What do you think?
Thanks.
Max
Max
Regards
-- Daniele Furlan