Thanks for your response.

> Beacon and RANN action frame are both generated by the mac80211. So
> for adding vendor specific information, you probably need to look into
> the mac80211.
I will look into the mac80211 subsystem in order to accommodate the
missing pieces of information that are required. It might also make
more sense to have it in order to maintain compatibility with the o11s
way of RANN broadcasting.

> G1 and G2, mesh gate are both set to same channel or frequency? If it
> is same channel, the radio itself is half duplex and also due to
> CSMA/CA so only, only one is able to Tx at one time. How much
> bandwidth gain that you achieve? What's the chipset are you using?

I checked the channels by setting up the experiment again. Using
wireshark and a changing the channel to all possible values of a
monitor interface (iw dev mon0 set channel X), it was evident that the
broadcast beacons from all the 4 cards were available on channel 1 and
channel 2. Also, the whole TCP/HTTP traffic was only visible on
channel 1 on the monitor.
Bandwidth gain seems to be linear when we start with a smaller link
speed (100 kbps) and then goes on to be somewhat constant once we
reach a higher link speed (800 kbps). Since we are using the Click
Modular router, our scripts go up till 1.4 MB/s of data transfer as
compared to the vanilla o11s' 2.5 MB/s using the same hardware.

The experiment was carried out using one TP LINK (TL WN722N HIGH GAIN)
and three TL-WN721N cards.
Both these models are based on the Atheros AR9002U chipset.

Regarding the gains, we've also plotted a graph, which you can find
here, 
https://github.com/scar1337/nodeman/blob/master/experiment/experiment_tcp/ExperimentGraph.png
That graph should give you some idea about the gains.

> Is this similar to link aggregation using bonding
> (http://linux-ip.net/html/ether-bonding.html#ether-bonding-aggregation)
> that you want to achieve?

Since you suggested channel bonding, I actually looked into it. It
does seem similar. I was under the impression that channel bonding is
mostly used for fault tolerance and increasing reliability (similar to
a dual WAN router). But it seems it also is used for increased
throughput. I think what we are trying to achieve is effectively an
intelligent form of channel bonding. Hoping to hear any
suggestions/feedback on this, so this project can be continued :)

Regards,
Ashish Gupta

>
> ---
> Chun-Yeow
_______________________________________________
Devel mailing list
[email protected]
http://lists.open80211s.org/cgi-bin/mailman/listinfo/devel

Reply via email to