I recently spent several days figuring this out, so I thought I would
report my findings to the community in case anyone else is having the
same problem.
For one of my sites Analog has been reporting total bytes transferred
numbers of several times what the router is reporting, larger than the
Internet connection is capable of in fact. The affected site has several
large (>100 Meg) downloads, but is otherwise "normal".
It turns out that Apache does not report bytes sent, but instead reports
the number of bytes that "could have been" sent. In most situations
these numbers are the same, but in certain cases they can be very
dramatically different. On this particular site, Apache is reporting
*ten* times the data transferred then is actually going across the wire.
The crucial case is caused by the combination of very large downloads
(>100 Meg in my case, but anything over 1 Meg is probably being affected
to some extent) and particular download managers on Windows. Many
download managers will start several transfers at the same time, each
starting from a different place in the file. When one transfer gets to a
point covered by another transfer the download manager will abort the
transfer. When this happens Apache will report the number of bytes from
the start of the transfer to the end of the file, but the actual
transfer will be much shorter (because it was aborted).
This situation is most common with 206 Partial Transfer status codes,
but can also affect regular requests. The Apache developers are aware of
this "issue", but don't consider it a bug. Instead they developed the
logio module to report actual numbers of bytes transferred. If you are
using Apache, have large downloads on your site, and care about accurate
byte counts in Analog I suggest you start using the logio module and use
the %O format specifier instead of %b in your log formats.
There exists a download manager which will start a transfer for every 1
Meg of file size. With a 100 Meg file Apache will report 100 requests
with a total size of about 5 Gig! Meanwhile the actual data transferred
will be only slightly more than 100 Meg, a difference of nearly 5000%!
Other download managers have less dramatic behaviors, but still
routinely result in 300% to 700% differences between reported and actual
byte counts on large downloads.
Jason
--
[EMAIL PROTECTED]
--
Dr. Seuss books . . . can be read and enjoyed on several levels. For
example, 'One Fish Two Fish, Red Fish Blue Fish' can be deconstructed
as a searing indictment of the narrow-minded binary counting system.
-- Peter van der Linden, Expert C Programming, Deep C Secrets
+------------------------------------------------------------------------
| TO UNSUBSCRIBE from this list:
| http://lists.meer.net/mailman/listinfo/analog-help
|
| Analog Documentation: http://analog.cx/docs/Readme.html
| List archives: http://www.analog.cx/docs/mailing.html#listarchives
| Usenet version: news://news.gmane.org/gmane.comp.web.analog.general
+------------------------------------------------------------------------
