|
|
|
|
|
|
|
|
|
|
|
| Re: Eclipse Download Statistics [message #16004 is a reply to message #15986] |
Fri, 25 February 2005 15:32   |
 Eclipse Webmaster
Messages: 607353
Registered: July 2009 |
Senior Member |
|
|
The stats come from a MySQL database. When you download a file that
prompts you to select a mirror, I track the file, the mirror you
selected, and test the mirror for the presence of the file, into a MySQL
database. This way, the numbers you see include downloads from mirror
sites as well as from eclipse.org.
Parsing the Apache logs is tedious for any application because of the
volume of logs generated (eclipse downloads can generate in excess of 15
MB per day), and because it include other files, such as .jpg, .html and
..php files, it is rather inefficient. Also, our Apache logs do not
include mirror downloads, so they are inaccurate. The tracking via
MySQL is extremely clean and efficient, and very easy to query.
Denis
Sam Mesh wrote:
> Thanks Denis,
>
> Quite Impressive Numbers (not as ours:)!
>
> BTW, I was asking in BIRT for WebLog reporting.
> They agree this would be a good use case for BIRT.
>
> Any hints, where this statistics came from?
> What about mirrors?
>
--
Eclipse WebMaster - webmaster@eclipse.org
Questions? Consult the FAQ at http://www.eclipse.org/webmaster/faq.html
View my status at http://www.eclipse.org/webmaster/main.html
|
|
|
|
| Re: Eclipse Download Statistics [message #16040 is a reply to message #16004] |
Fri, 25 February 2005 19:36 |
Eclipse User |
|
|
|
Originally posted by: SamMesh.gmail.com
As Ian mentioned, "So the figures show requested downloads, not
necessarily completed ones." AFAIK, from Apache logs we can determine
that download completed?
Of course, mirrors are out of the scope of detailed logging unless ...?
BTW, we have about one mega-logo-day of Apache logs and still can do a
full year report from Web Log Expert (hope to replace them with BIRT). I
think, BIRT will do comprehensive reporting of the 15 mega-logo-days
importing raw log data into something well structured. Personally, will
help them. :)
--
Sam Mesh
WebMaster wrote:
> The stats come from a MySQL database. When you download a file that
> prompts you to select a mirror, I track the file, the mirror you
> selected, and test the mirror for the presence of the file, into a MySQL
> database. This way, the numbers you see include downloads from mirror
> sites as well as from eclipse.org.
>
> Parsing the Apache logs is tedious for any application because of the
> volume of logs generated (eclipse downloads can generate in excess of 15
> MB per day), and because it include other files, such as .jpg, .html and
> .php files, it is rather inefficient. Also, our Apache logs do not
> include mirror downloads, so they are inaccurate. The tracking via
> MySQL is extremely clean and efficient, and very easy to query.
|
|
|
|
| Re: Eclipse Download Statistics [message #16127 is a reply to message #16077] |
Mon, 28 February 2005 13:03 |
| Eclipse User |
|
|
|
Originally posted by: ian.no.spam.please
Eclipse WebMaster (Denis Roy) wrote:
> Ian Darwin wrote:
>
>> So the figures show requested downloads, not necessarily competed ones.
>
>
> Impossible to tell with any degree of accuracy if a download completes.
> Especially if it was served by a mirror. One could refine the stats to
> eliminate duplicates based on IP and file, but hey. This data isn't
> going to be used for patenting a cure for cancer, right?
Sure, I wasn't trying to start an argument, just comment on the validity
of the statistical data. I agree the figures are impressive!
|
|
|
Powered by
FUDForum. Page generated in 0.05753 seconds
Home