wireless environment by Mininet and NS3

[Farzaneh Pakzad]

Hi,

I am researching about SDN and how we can use it in wireless networks. 
So I found this work in this area in which it is integrated ns3 with Mininet and did something very special.

https://github.com/mininet/mininet/wiki/Link-modeling-using-ns-3

I walked through the tutorial and also did most of the examples in Mininet. 

Finally,  I made a linear topology with multiple switches connecting together via WiFiBridgeLink().(As mentioned in the document we can only use this type of link to connect the switches together.
 Which means for each new link connected to the wireless switch we add another interface, instead of using one interface for one wireless switch)

And the first and last switches are also connect to the host via Mininet Link.

Actually I want to measure the throughput between this two hosts connected via switches.

As you know wireless links are half-duplex and if we want to measure the bandwidth between multiple hops, by increasing the number of the hops the bandwidth would decrease.

I did this experiment with a hosts only network and adding WiFiSegment() to each node and it worked properly but when I used switches and WiFiBridgeLink() it acts like a wired link.

I attached a graph of my experiment in which I set the OFDM rate to 6 Mbps, and as you can see for the network with switches I have somehow constant throughput, but for the host only network  the throughput would decrease by increasing the number of hops.

In reality the switches should also work in a same way as hosts works but  here they didn't!!! 

Would you please help me about that?

Is there any way that we can implement wireless network with switches with one radio interface instead of using multiple ones?

I highly appreciated any ideas about that.


Regards,
Farzaneh.

[Tommaso Pecorella]

Hi,

I have no experience about this, but it looks like a collisions effect.
Double check that in the two simulated systems (host and switch) the links in each node are using orthogonal channels (and extend this check to the nodes in range).

Please pay attention that YansWifi will not consider adjacent channels interferences AND if you do a mistake in creating the channels, also in the same channel.

In other terms, it is possible that the "switch" configuration is highly optimistic.

Cheers,

T.

[Farzaneh Pakzad]

Hi,

Thanks for the reply and explanation.

I set the channel to  use OFDM.
But the issue is using multiple interfaces on a siwtch.
 I noticed that you mentioned " YansWifi will not consider adjacent channels interference" . Would you please let me know which model could I use instead to consider it ?

I also agree with the term optimistic about the switches.

Best regards,
Fazaneh

[Tommaso Pecorella]

The new SpectrumWifiPhy is slower than YansWifiPhy but it does take into account adjacent channels interferences.

Cheers,

T.