Skip to first unread message
feffir...@gmail.com
Sep 5, 2016, 8:12:10 PM9/5/16
to ns-3-users
Hi,
I have connected two Linux Containers (left and right, where left owns two veth interfaces) via ns-3 in wifi adhoc mode. Everthing seemed to work well, until I found out that I am losing some packets at the TapBridges. The ns-3 script I am using is basically a slightly modified version of tap-wifi-virtual-machine.cc from src/tap-bridge/examples. To create the containers, bridges and tap devices I followed the HOWTO: https://www.nsnam.org/wiki/HOWTO_Use_Linux_Containers_to_set_up_virtual_networks
Inside the containers, I am running iperf3 for bandwidth estimation. If I listen to tap-left (where the iperf3 client is located) with Wireshark and compare the number of UDP packets received with those from the pcap log from the corresponding node in ns-3, there is a difference of about 200 packets (I repeated this several times). There is no loss on the other side (tap-right). This is already a huge improvement due to the patch from here: http://wiki.confine-project.eu/soft:ns3-with-vct
Otherwise, there were only about 3% of the send packets received at the other end, which was quite a catastrophe.
So where are my last missing packets? It has to have something to do with the combination of the TapBridge and wifi. If I run the example script for CSMA (and use "UseLocal" as mode like in the wifi script), there is absolutely no loss.
Since this is my first time working with lxc, tap devices and ns3::TapBridge I am a little bit lost at this point. I would really appreciate a solution to this problem or a useful hint at what to investigate next.
Thanks a lot,
Stefanie
PS: I attached the shell script for setting up the taps and bridges and the ns-3 script, which I use to connect the containers. I am using Ubuntu 14.04 (64bit) and the latest version of ns-3-dev (tried to update with mercurial today, but nothing changed).
I have connected two Linux Containers (left and right, where left owns two veth interfaces) via ns-3 in wifi adhoc mode. Everthing seemed to work well, until I found out that I am losing some packets at the TapBridges. The ns-3 script I am using is basically a slightly modified version of tap-wifi-virtual-machine.cc from src/tap-bridge/examples. To create the containers, bridges and tap devices I followed the HOWTO: https://www.nsnam.org/wiki/HOWTO_Use_Linux_Containers_to_set_up_virtual_networks
Inside the containers, I am running iperf3 for bandwidth estimation. If I listen to tap-left (where the iperf3 client is located) with Wireshark and compare the number of UDP packets received with those from the pcap log from the corresponding node in ns-3, there is a difference of about 200 packets (I repeated this several times). There is no loss on the other side (tap-right). This is already a huge improvement due to the patch from here: http://wiki.confine-project.eu/soft:ns3-with-vct
Otherwise, there were only about 3% of the send packets received at the other end, which was quite a catastrophe.
So where are my last missing packets? It has to have something to do with the combination of the TapBridge and wifi. If I run the example script for CSMA (and use "UseLocal" as mode like in the wifi script), there is absolutely no loss.
Since this is my first time working with lxc, tap devices and ns3::TapBridge I am a little bit lost at this point. I would really appreciate a solution to this problem or a useful hint at what to investigate next.
Thanks a lot,
Stefanie
PS: I attached the shell script for setting up the taps and bridges and the ns-3 script, which I use to connect the containers. I am using Ubuntu 14.04 (64bit) and the latest version of ns-3-dev (tried to update with mercurial today, but nothing changed).
Reply all
Reply to author
Forward
0 new messages