[Systems] [IAEP] Content Delivery Network is in production

[David Farning]

On Mon, Oct 12, 2009 at 11:55 AM, Yama Ploskonka <yamaplos at gmail.com> wrote:
> I'll post to Sur.  Pablo Flores might be the prime actor here.
>
> Dunno if 'guay can ever use the .86, the whole point of .84 was to make it
> usable by the guayish security system AFAIK, but I am suitably ignorant of
> those matters as to fact.

Yes, this is one of the challenges of upstream, downstream
relationships.  Upstreams release according to the mantra "early and
often."  Downstreams deploy according to the mantra "Seldom and
stable."

Eventually, we will get this worked out.  The big thing is for:
1. Sugar Labs to focus on innovating.
2. OLPC (and other oems) to focus on stabilizing those innovations
into a series of hardened products.
3. Deployments to focus on deploying and supporting their customer base.

> If I were right, and I am afraid I might be :-) then we will solve nothing
> on the OLPC servers getting overworked, that will stay, but we might be able
> to get the bundles to be better accessible, but who knows.  By now I've lost
> track who handles servers for OLPC.  Could someone copy direct?

I always go through the Volunteer Infrastructure Group when working
with OLPC on infrastructure stuff.

david

> On Mon, Oct 12, 2009 at 11:10 AM, David Farning <dfarning at sugarlabs.org>
> wrote:
>>
>> On Mon, Oct 12, 2009 at 10:56 AM, Yama Ploskonka <yamaplos at gmail.com>
>> wrote:
>> > wow wow wow, this is a way cool concept.  Local updates that would be
>> > truly
>> > local automagically.  I really would want to know how this ends up
>> > working
>> > for the large deploys , esp. Uruguay.  I have been told that some
>> > servers
>> > are quite overstressed already as it seems (correct me here please) that
>> > as
>> > the machines there are booted they automatically try to connect
>> > somewhere
>> > OLPCish.  If it were true, I guess that should be fixed.
>>
>> Some where between 80 and 90% of activity bundle download traffic
>> comes from Uruguary:)  I am hoping to find a couple of mirrors in
>> Uruguay in the next couple of days.  If you have any leads....  We
>> need room for about 30GB of content.
>>
>> Yes, .82 deployments are set to communicate directly with olpc
>> servers.  This has been updated in .86.  Another option is for Sugar
>> Labs to work with the VIG to make our mirror system available to OLPC.
>>
>> david
>>
>> > On Mon, Oct 12, 2009 at 9:34 AM, David Farning <dfarning at sugarlabs.org>
>> > wrote:
>> >>
>> >> > Seems to be working pretty well from here in Portugal, download.sl.o
>> >> > gets redirected to download2.sl.o (solarsail @ MIT?) and utwente.nl.
>> >> > Both max out my connection (1.2 MB/s).
>> >>
>> >> Nice, we will be adding mirrors as the system stabilizes.  The goal is
>> >> for the download infrastructure to  be ready to handle 10X the
>> >> bandwidth usage as on the last soas release.
>> >>
>> >> As soon as then SoaS and Marketing teams decide dates for the release
>> >> and associated press releases can someone send notices of those dates
>> >> to the systems mailing.
>> >>
>> >> Last year I had to restart the web server on sunjammer every couple of
>> >> hours when it froze:(  It would be nice to have Bernie and the other
>> >> infrastructure gurus around on days we expect larger than normal loads
>> >>
>> >>
>> >> david
>> >> > Congrats,
>> >> >
>> >> > Tomeu
>> >> _______________________________________________
>> >> IAEP -- It's An Education Project (not a laptop project!)
>> >> IAEP at lists.sugarlabs.org
>> >> http://lists.sugarlabs.org/listinfo/iaep
>> >
>> >
>
>