[Sugar-devel] ESSIDs and BSSIDs, NM and Sugar

[Andrés Nacelle]

Hello there,
I´m writing you from Plan Ceibal - Uruguay. Here I´ve been doing some test
on this subject trying to see what was happening with this "jumping" of the
XO from one to another AP.

Basically what I did was configure two AP with identical SSID, so the XO
would see them both under the same balloon. The AP´s were distribute in the
testing room so that I would have basically 4 regions:

   1. one place were the signal strength seen by an XO from both AP was the
same (-40 dBm).
   2. the same as before but with lower str (-52dBm).
   3. one place with signal isolation of 14dBm (one signal at 42dBm and the
other one at 56dBm).
   4. one place with signal isolation of 40dBm (one signal at 18dBm and the
other one at 58dBm).

After having the scenario prepared I put 60 XO (with 801) making some
traffic to the server and start monitoring, on one XO in each region, the
MAC (from the AP) to witch they were associated.

Basically the results in this was that in regions 1, 2 and 3 the jumping was
about 50 % of the time, staying something between 30 seconds to 20 min in
the same AP. In region 4, 90% of the time the XO stayed on the same AP.

In other words I agree on you with the no smart dynamic selection on terms
of quality of signal by the XO. I really don´t know yet the reason but it
happens.

I didn´t check the blocks in the ~/,sugar/default/nm/networks.cfg, but as
soon as I can run some more tests on this I´ll keep an eye on this.

I wasn´t able to find data witch allow me to say if the throughput was
affected because of this behavior (I suspect that would be some decrease on
congested nets), but I don´t like to have in a school half of the XO moving
from an AP to another all the time. This kind of behavior I think would
produce dynamical congestion in the net, basically because when many XO
jumps to the same AP this is not going to be able to manage all that much
connections and throughput, then the XO are going to keep trying and
suddenly they will start connecting to the other. What I´m afraid of is a
mass effect, in witch big blocks of XO go from one to another AP
sub-employing one of the AP and overcharging the other.

We expect to have more chances to study this problems, but now we need to
move on, so meanwhile we are planing to use different SSID for different
channels and  keep a design in witch equal channels don´t step on each
other.

Hope this help the rest of you on your own tests

If there is any suggestion on how to fix this, I will gladly run the tests
and try it but wright now we are not sure were to search for this solution.

Keep up the spirit that a solution may appear for this.

Bye

Andres Nacelle
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.sugarlabs.org/archive/sugar-devel/attachments/20090810/05b6b6ff/attachment-0001.htm