[Sugar-devel] feedback from a teacher in Uruguay

David Van Assche dvanassche at gmail.com
Fri Aug 14 13:22:00 EDT 2009


Yeah, with that many clients, I'd play with the s2s option to load balance
the xmpp traffic. Either that, and/or take out all debugging options, error
feedback, and the like, so you only get the needed traffic... This page
gives some tuning options:
http://www.ejabberd.im/tuning

Enabling Gadget is supposed to really help, too. Please let us know how it
all goes, as this kind of information is invaluable. From the realtively few
reports we get about collaboration scaling, it seems xmpp based sharing is
not very widespread yet. I guess this is fortunate in one way (things
generally work), but unfortunate in another (not enough feedback)

Also this from ejabberd site:

"Kernel poll reduces ejabberd's CPU usage when it has hundreds (or more)
network connections. It does not affect memory consumption or latency, so if
you plan to support great amounts of simultaneous connected users, you
really want to have a Jabber server with Kernel Poll support."
There is also the possibility of using a different server, but from
everything mentioned, it does seem like Erlang based ejabberd will scale
better than anything else. But good things have been said about Java based
Openfire and Tigase, and Lua based Prosody. We are actually doing an
internal evaluation of the various xmpp servers at my work, so I'll try
publishing that...

good luck,
David Van Assche

On Fri, Aug 14, 2009 at 5:49 PM, Andrés Nacelle <anacelle at plan.ceibal.edu.uy
> wrote:

> Hi David, we're aware of this situation. In fact we attempt to share
> activities between this mush XO in order to compare the performance against
> the ejabber. We've got good results with the ejabber, but we don't know what
> will happen in a big school where 200 or 300 XO may appear n the
> Neighbourhood view.
> Without the ejabber lots of udp traffic is generated, from which about 60%
> are retransmissions. If ejabber is used you still should get an improvement
> on the performance if you disable the beacons and the probe response, they
> can produce saturation in the net for some seconds.
>
> Thank for you advise and help.
>
> Andres Nacelle
>
> 2009/8/14 David Van Assche <dvanassche at gmail.com>
>
> Hi Andrés,
>>   The sharing of activities not working for more than a few connections is
>> a common issue in all the deployments I have seen. It is really important to
>> have an ejabberd server to help with routing xmpp traffic, if you plan on
>> running more than a few XOs in collaborative mode.
>>
>> kind regards,
>> David Van Assche
>>
>>
>> On Fri, Aug 14, 2009 at 3:14 PM, Andrés Nacelle <
>> anacelle at plan.ceibal.edu.uy> wrote:
>>
>>> Hello you all,
>>>
>>> In first place thanks for asking, 'm glad to try to help here.
>>>
>>> Well, lets see. I think in first place we need to make sure witch is the
>>> version of the SO and firmware on the XO from the teacher and children,
>>> which is really easy for the teacher to tell us. There's a chance that she
>>> (or some student) have an activity called something like "Info XO", in that
>>> case they just have to run it and the info will be displayed. In case she
>>> doesn't have it she needs to pres the "tick" button on the game pad during
>>> the booting for the firmware and after the booting change to a terminal an
>>> there look for the SO version.
>>> Once we have that info we can start to see if we can do something about
>>> the mesh issue.
>>> Independent of this, we never been able to share activities on the mesh
>>> for more than a few XO, trying with 20 no XO shared a thing, not even one.
>>>
>>> On the other topics I have nothing to say different from what has been
>>> said in the mailing list.
>>>
>>> Thank
>>>
>>> Andres Nacelle
>>>
>>>
>>>
>>>
>>> 2009/8/14 Tomeu Vizoso <tomeu at sugarlabs.org>
>>>
>>> 2009/8/13 Gabriel Eirea <geirea at gmail.com>:
>>>> > Hi all,
>>>> >
>>>> > I'm sending below the original text (in Spanish) of a teacher friend's
>>>> > request. I'll try to translate her main concerns.
>>>> >
>>>> > 1) Mesh network doesn't work. Very few kids can get in the same
>>>> > "network". She claims it got worse with the latest update in Ceibal.
>>>>
>>>> Not sure we can debug this remotely :/
>>>>
>>>> Andrés, do you know about any work in the Ceibal plan to determine the
>>>> reliability of the mesh network?
>>>>
>>>> > 2) She saves Write documents in the pendrive, then she opens them at
>>>> > home with KUbuntu and OpenOffice, then she saves the modified document
>>>> > in the pendrive and tries to open it in the XO with Write. She claims
>>>> > she can't open it even if the document was not modified.
>>>> >
>>>> > 3) Same scenario as 2), she saves it from KUbuntu as a pdf in a
>>>> > pendrive. She can't open it back in the XO (she gets a "blank page").
>>>> >
>>>> > 4) The previous cases are frustrating because she wants to be able to
>>>> > move back and forth between the XO and her (convenitonal) laptop.
>>>>
>>>> I understand the frustration, do you think we could get copies of
>>>> those files and more details about the workflow?
>>>>
>>>> About usb sticks, in versions earlier than 0.84 we are using a scheme
>>>> that is quite fragile, specially when using the same stick with
>>>> non-sugar computers.
>>>>
>>>> I would recommend deleting the .olpc.store in the usb stick after
>>>> every usage in a non-sugar computer and see if weird things stop
>>>> happening. If this helps somehow, then we should find a way to
>>>> publicize this trick. Note that from Sugar 0.84 this is not needed.
>>>>
>>>> > 5) SocialCalc save and restore doesn't work. After saving, there is a
>>>> > Journal entry but when they open it the information is not there.
>>>>
>>>> Manu, know anything about it? I guess having a ticket with the
>>>> activity logs would help here.
>>>>
>>>> > 6) She claims that files dissappear for no reason. This probably needs
>>>> > to be clarified. She discards the possibility of kids deleting files
>>>> > by accident.
>>>>
>>>> Yes, this deserves further clarification. It's a very wide issue, so
>>>> we'll need a way to track all the different circumstances in which
>>>> data reliability can happen. The journal data store was rewritten in
>>>> 0.84 with increasing reliability on mind, but I guess it will take a
>>>> while to update all machines to that or later versions of Sugar.
>>>>
>>>> > Now the question is, how can we get these questions in reasonable
>>>> > shape so it is useful information for debugging?
>>>>
>>>> That's the Million Laptops Question ;)
>>>>
>>>> We are doing nice progress on this, but it's a really big task.
>>>>
>>>> Thanks a lot,
>>>>
>>>> Tomeu
>>>>
>>>> > Thanks,
>>>> >
>>>> > Gabriel
>>>> >
>>>> >
>>>> > ---------- Mensaje reenviado ----------
>>>> >
>>>> > Quisiera que supieras que tenemos algunos problemas con varias cosas
>>>> > en la xo y me parece que son puntos importantes para que podamos
>>>> > trabajar mejor con la máquina. te paso algunos, y vos ves si pueden
>>>> > hacer algo al respecto.
>>>> > Las redes malla siguen funcionando mal, y desde que se actualizaron
>>>> > las máquinas es peor, son muy pocos los niños que logran entrar en la
>>>> > misma red.
>>>> > He tenido dificultades cuando guardo en el pen drive textos realizados
>>>> > en write, los leo con el open office y luego vuelvo a cargarlo en sus
>>>> >  máquinas y no me los lee, aún cuando no les realizo ninguna
>>>> > modificación. Probé cambiándole el formato a pdf, que sé que las xo lo
>>>> > leen, y no pude, cuando abro el pen en la xo y hago clic en el
>>>> > archivo, ahí puedo ver que está el texto (no lo abrió aún) y hago clic
>>>> > en retomar y qqueda la hoja en blanco. no tengo idea como voy  a hacer
>>>> > para que reparen el texto si ellos no lo bajan en la xo.
>>>> > yo no entiendo, pensé que al tener kubuntu y el open office podría
>>>> > trabajar igual que los niños, pasando de una máquina a la otra sin
>>>> > problemas. Necesito saber más para poder manejarme mejor.
>>>> > otra dificultad que encontré es que en el social calc, la planilla
>>>> > electrónica, a la mitad de los chicos se les borra la información al
>>>> > guardar, es decir, queda en el journal como guardada, pero los datos
>>>> > no aparecen en pantalla cuando abrís.
>>>> > Pasa mucho que les desaparecen los archivos, se les borran solos.
>>>> > Antes creía que eran ellos que los borraban sin querer, pero ahora me
>>>> > doy cuenta que es así.
>>>> > _______________________________________________
>>>> > Sugar-devel mailing list
>>>> > Sugar-devel at lists.sugarlabs.org
>>>> > http://lists.sugarlabs.org/listinfo/sugar-devel
>>>> >
>>>>
>>>>
>>>>
>>>> --
>>>> «Sugar Labs is anyone who participates in improving and using Sugar.
>>>> What Sugar Labs does is determined by the participants.» - David
>>>> Farning
>>>>
>>>
>>>
>>> _______________________________________________
>>> Sugar-devel mailing list
>>> Sugar-devel at lists.sugarlabs.org
>>> http://lists.sugarlabs.org/listinfo/sugar-devel
>>>
>>>
>>
>>
>> --
>>
>> Joan Crawford<http://www.brainyquote.com/quotes/authors/j/joan_crawford.html> - "I, Joan Crawford, I believe in the dollar. Everything I earn, I spend."
>
>
>


-- 

Ogden Nash <http://www.brainyquote.com/quotes/authors/o/ogden_nash.html>  -
"The trouble with a kitten is that when it grows up, it's always a cat."
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.sugarlabs.org/archive/sugar-devel/attachments/20090814/f8744aeb/attachment-0001.htm 


More information about the Sugar-devel mailing list