On 2 December 2014 at 11:27, Owen Taylor <otaylor@redhat.com> wrote:
On Tue, 2014-12-02 at 11:12 -0700, Stephen John Smoogen wrote:

> On 2 December 2014 at 09:22, Matthew Miller <mattdm@fedoraproject.org>
> wrote:
>         On Tue, Dec 02, 2014 at 05:15:38PM +0100, jfm512@free.fr
>         wrote:
>         > Now it seems Fedora's downloads have been steadily going
>         downwards
>         > and it is highly improbable that is due to a spectacular
>         increase of
>         > number of people setting local mirrors.
>         Smooge is working on some updated statistics, and from what
>         I've seen,
>         it's mostly flat, but not actually steadily downwards. More on
>         the
>         other things later. :)
> Temporary URL for picture of things from 2007 til last month:
> https://smooge.fedorapeople.org/stats-2014/
> I am hoping to have a permanent url and updated set by next month.

Is there some explanation why there the Fedora vs. Epel split in the
last graph so closely tracks the x86_32 vs. x86_64 split in the second
graph, as if all EPEL users were on x86_64 and all Fedora users were on

OK going over the raw data for 2014-10.. it is about

83.3% of EPEL users are x86_64 
16.6% of EPEL users are i386
00.04% of EPEL users are ppc
00.06% of EPEL users have misconfigured yum

66.6% of Fedora users are x86_64
33.0% of Fedora users are i386
00.2% of Fedora users are arm
00.08% of Fedora users are ppc
00.02% of Fedora users are either unknown or s390/aarch64/etc

I will try and separate out the Fedora only architecture graphs this week.

Caveats are that these are people directly interfacing with mirrormanager versus being configured to go against an internal mirror. 

Stephen J Smoogen.