Re: [Summary-Talk] Variations in page counts

[Jason Linhart <jason@xxxxxxxxxxx>]

I wouldn't expect the numbers to be as different as what you describe.
There will almost always be some difference, but it should be smaller
(usually two or there and almost always less than 50) and more
consistent from page to page. The differences you are seeing might be an
indication of some kind of problem with your server.

Report 10 - Page Requests is showing the number of successful requests
to each page. Reports 100/110/120 - Daily/Weekly/Monthly Pages are
showing the number of requests, successful or not, minus one less than
the "Hit threshold for time history reports" (on the Memory
Configuration page).

The "Hit threshold for time history reports" is normally a small number,
two or three, though some people set it much higher, which is the same
for every page and can usually be ignored. If it bothers you you can
always set it to zero, though that takes more memory.

The difference between successful requests and all requests is the
number of errors for that item. On most servers, where the content has
existed for the entire time period in question, there won't be any
errors. So in the common case these numbers are the same. However some
servers will once in a while fail to serve content even though it
exists, resulting in detectable numbers of errors. This typically
happens when the server is overloaded, but can also be caused by
improperly formed HTTP requests.

In theory these numbers can also be different if you have "Expire req.
and ref. if not hit every N days (0-don't)" turned on, but in practice
requests that are getting expired will almost never show up in the
Daily/Weekly/Monthly reports at all.

My guess is that your server is failing to serve the page about 1% of
the time. This could be because of the server getting overloaded, or it
might be invalid HTTP requests (though I have never seen more than a
couple of those a day), or it could be caused by pages that are large
enough that when the end user aborts the page load (for example by
hitting the back button) the page hasn't already completed sending from
the servers point of view and an abort is logged. Looking at the Details
: Log Fields : Result Code report and seeing which error codes, other
than 404, are common should give you some hints about what might be
happening.

Jason


Jeff Whitmill wrote:
> A media client of ours sends the following report and questions:
> 
>>I noticed an anomaly between the total pages numbers for the year on 
>>oursite.org/statistics/december2005/report-10.html and the monthly 
>>pages numbers on oursite.org/statistics/december2005/report-120.html. 
>>
>>For the home page, it's a small difference:  770,785 on the monthly, 
>>cf. 769,243 on the total; a diff of 1,542 more on the monthly.
>>
>>But for the TV Schedules page, the difference is more pronounced:  
>>427,932 on the monthly cf. 423,980 on the total; a diff of 3,952 more 
>>on the monthly - almost a 1% difference.
> 
>>The weekly and daily pages are showing the same year totals as the 
>>monthly pages.  That kinda tosses out the theory that the data degrade 
>>as you get more granular.  What do you think?  Why would we be getting 
>>that much, and that inconsistent, of a variation?
>>
>>Or I could just leave it be and take it for what it is and not worry 
>>about it?
> 
> What can I tell her about this?

-- 
Jason@Summary.Net
--
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

-- 
Jason@Summary.Net
--
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

-------------
Go to <http://summary.net/list.html> to update subscription info.