Is there any polling mechanism in wmediumd.c ？

[zijinchen]

Hello, I have checked your wmediumd.c file and found the CSMA code, but I did not find the polling code, I wonder if there is any?Thank you very much!

[ramonfontes]

Hi,

I haven't finished the coding yet.

[zijinchen]

@ramonfontes
Thank you very much for your reply! Because I see that the mininet-wifi project uses wmediumd on your version, and I want to confirm that you mean that there is no polling function in the current code, right? Do you plan to add it later? Or is it not necessary? Thank you very much!

[ramonfontes]

Are you refering to this repo https://github.com/ramonfontes/wmediumd?

[zijinchen]

@ramonfontes
What I see is this：https://github.com/ramonfontes/wmediumd_802154
I haven't noticed the one you are referring to. Could you briefly tell me the difference between the two projects?

[ramonfontes]

Wmediumd was designed for 802.11 and I was trying to extend it to 802.15.4

[zijinchen]

@ramonfontes
Great job! I know that there is CSMA in 802.11, but I am not sure whether there is a polling mechanism in 802.11.Is there any reflection in your two projects?thank you！

[ramonfontes]

It has a backoff time https://github.com/ramonfontes/wmediumd/blob/mininet-wifi/wmediumd/wmediumd.c#L384 that I'm not pretty sure that is working correctly. I would say that today's biggest problem with wmediumd is the separation of the wireless medium. For example, data traffic exchanged between 2 nodes associated with one access point directly interferes with simultaneous traffic between 2 other nodes associated with another access point.

[zijinchen]

@ramonfontes
You are absolutely right. This is the second question I want to ask, because I see that the four configurations you mentioned don't take into account the interference between signals, for example, different signals on the same channel? I don't know if this description is the same as what you said about the biggest problem?

[ramonfontes]

Yes, they don't take into account the interference between two signals on the same channel too. However, even if you add two nodes associated with one AP and other two nodes associated to another AP (far from the first three nodes), it doesn't work well. I started to work with this 15 days ago with https://github.com/ramonfontes/wmediumd/tree/parallel, but I'm not sure whether this is the correct direction.

[zijinchen]

@ramonfontes
However, the interference between multiple signals and multiple aps are more in line with the actual situation. What you mean is that even if two aps (with some nodes) are far away from each other and there is no signal interference, the effect is not very good, right?So far only a single AP can be satisfied?

[ramonfontes]

You can see the data being exchanged through the 2 APs. The problem is: if you try Iperf between the first 2 clients and throughput is 10Mbp/s; and simultaneously do the same between the other 2 clients, the throughput will be split among all nodes, resulting in lower throughput.

[zijinchen]

Sorry, I am a little confused. According to your description, the throughput of the first two clients is set to 10Mbp/s, and the throughput of the last two clients is also set to 10Mbp/s, so the throughput will decrease. Is the throughput of the whole system fixed?The more AP, the less throughput per allocation?

[ramonfontes]

The throughput decreases after starting the second instance of iperf, which is being used with nodes that aren't associated with the same AP. I mean, the second instance interferes the results of the first one (and vice-versa).

[zijinchen]

Aha! isn't that the "distraction" we want? Is it strange that you set the distance between the two AP's far? There should be no interference in theory, but there is interference in practice, Right?

[ramonfontes]

No, we don't want this when there is no signal overlap.

[zijinchen]

The first sentence was just a joke, I know what you mean, the signals should not interfere with each other without overlapping, but in fact the experiment you did, the two far away from the AP caused interference, now there is a problem, right?

[ramonfontes]

Okay! :) well, I suppose there are two big problems now. ;)

By the way, are you trying to fix the problem mentioned by you?

[zijinchen]

Yes, we are also committed to solving this big problem. We hope to communicate with you more. Could you please keep one email address?

[ramonfontes]

Yes, sure. Please let me know if you need help or discuss more about this. There are many strands about mininet-wifi and I have not been able to handle it all. :|

We could have done all this discussion on /wmediumd

[email protected]