[Server-devel] XS server testing

tkkang at nurturingasia.com tkkang at nurturingasia.com
Fri May 29 10:20:29 EDT 2009


Thanks for the great reply and enlightenment. It is timely as we are just rounding up people for tommorrow. I will pass the message to the team and see how we approach it.

Will have a long domain name:  schoolserver.schoolname.example.org

I hope this is OK.


>-----Original Message-----
>From: Daniel Drake [mailto:dsd at laptop.org]
>Sent: Friday, May 29, 2009 10:03 PM
>To: tkkang at nurturingasia.com
>Cc: 'Martin Langhoff', 'Reuben K. Caron', 'XS Devel'
>Subject: Re: [Server-devel] XS server testing
>
>2009/5/28  <tkkang at nurturingasia.com>:
>> At the moment we are concern about the wireless support. A message below from a volunter:
>>
>> --
>> Yesterday in TK's office, I experienced that registration takes only 2 second for one XO, so I guess the bandwidth requirement for registratoin process is very minimal.
>
>Yes, the registration is a simple process that does not use much bandwidth.
>
>> In my viewpoint, the most important question in the Sat. stress test is whether we should use 1 AP or 3 AP. If it is feasible to register 50 PC via one AP simultaneously, then we are safe. Because it means that we can register 150 XO in chanenls 1, 6,11 simultaneously in SiChuan school. Then we can finish registering 450 XO in 3 rounds. What do you think?
>
>Given that registration is a light process and that it would actually
>be quite a logistical challenge to get 50 people to click "Register"
>in exactly the same moment, I guess your concerns are more about
>general prolonged usage of the network, rather than specifically the 1
>second that is needed to register a laptop.
>
>You may have problems getting 50 people connected to the same access
>point (the step *before* registration). Some access points that I
>tested would crash after accepting the 31st user. Some are OK though.
>But in any situation, 50 is probably the maximum you'd want to put on
>1 access point.
>
>You may want to focus on channels 6 and 11, because "out of the box"
>XOs which have not yet connected to an access point will start meshing
>on channel 1, this may saturate the spectrum with a sufficient number
>of laptops.
>
>I do not think you need "rounds" of registration. With sufficient
>access points, you should be able to get 450 users online
>concurrently. Having insufficient access points will be a problem
>because the XOs automatically connect to familiar networks, so it will
>be logistically difficult to get the first round of children to
>disconnect to allow for the second round.
>
>And as for post-registration... As far as I can tell, the XS has not
>seen much field use with such a large number of users concurrently
>signed on ejabberd, so your feedback will be very useful. We have
>schools of similar size here in the Paraguayan deployment, but not
>everyone has been taught about registration and collaboration yet, so
>we're only at about 100 users online at any one time. (and actually
>ejabberd is not coping very well with that, see my emails about
>ejabberd CPU usage)
>
>Daniel
>




More information about the Server-devel mailing list