NCEP Home > NCO Home > Shared Infrastructure Services Branch > WWW/FTP Server Usage Statistics > About the Server Usage Statistics
About the Server Usage Statistics
The following information was taken from the documentation for The Webalizer. It goes into
the details concerning how the statistics are created and what they mean.
What is The Webalizer?
The Webalizer is a web server log file analysis program which produces
usage statistics in HTML format for viewing with a browser. The results
are presented in both columnar and graphical format, which facilitates
interpretation. Yearly, monthly, daily and hourly usage statistics are
presented, along with the ability to display usage by site, URL, referrer,
user agent (browser), search string, entry/exit page, username and country
(some information is only available if supported and present in the log
files being processed). Processed data may also be exported into most
database and spreadsheet programs that support tab delimited data formats.
The Webalizer supports CLF (common log format) log files, as well as
Combined log formats as defined by NCSA and others, and variations
of these which it attempts to handle intelligently. In addition, wu-ftpd
xferlog formatted logs and squid proxy logs are supported.
Gzip compressed logs may now be used as input directly. Any log filename
that ends with a '.gz' extension will be assumed to be in gzip format and
uncompressed on the fly as it is being read. In addition, the Webalizer
also supports DNS lookup capabilities if enabled at compile time. See
the file DNS.README for additional information.
This documentation applies to The Webalizer Version 2.01
The Webalizer produces several reports (html) and graphics for each
month processed. In addition, a summary page is generated for the
current and previous months (up to 12), a history file is created
and if incremental mode is used, the current month's processed data.
The exact location and names of these files can be changed using
configuration files and command line options. The files produced,
(default names) are:
index.html - Main summary page (extension may be changed)
usage.png - Yearly graph displayed on the main index page
usage_YYYYMM.html - Monthly summary page (extension may be changed)
usage_YYYYMM.png - Monthly usage graph for specified month/year
daily_usage_YYYYMM.png - Daily usage graph for specified month/year
hourly_usage_YYYYMM.png - Hourly usage graph for specified month/year
site_YYYYMM.html - All sites listing (if enabled)
url_YYYYMM.html - All urls listing (if enabled)
ref_YYYYMM.html - All referrers listing (if enabled)
agent_YYYYMM.html - All user agents listing (if enabled)
search_YYYYMM.html - All search strings listing (if enabled)
webalizer.hist - Previous month history (may be changed)
webalizer.current - Incremental Data (may be changed)
site_YYYYMM.tab - tab delimited sites file
url_YYYYMM.tab - tab delimited urls file
ref_YYYYMM.tab - tab delimited referrers file
agent_YYYYMM.tab - tab delimited user agents file
user_YYYYMM.tab - tab delimited usernames file
search_YYYYMM.tab - tab delimited search string file
The yearly (index) report shows statistics for a 12 month period, and
links to each month. The monthly report has detailed statistics for
that month with additional links to any URL's and referrers found.
The various totals shown are explained below.
Any request made to the server which is logged, is considered a 'hit'.
The requests can be for anything... html pages, graphic images, audio
files, CGI scripts, etc... Each valid line in the server log is
counted as a hit. This number represents the total number of requests
that were made to the server during the specified report period.
Some requests made to the server, require that the server then send
something back to the requesting client, such as a html page or graphic
image. When this happens, it is considered a 'file' and the files
total is incremented. The relationship between 'hits' and 'files' can
be thought of as 'incoming requests' and 'outgoing responses'.
Pages are, well, pages! Generally, any HTML document, or anything
that generates an HTML document, would be considered a page. This
does not include the other stuff that goes into a document, such as
graphic images, audio clips, etc... This number represents the number
of 'pages' requested only, and does not include the other 'stuff' that
is in the page. What actually constitutes a 'page' can vary from
server to server. The default action is to treat anything with the
extension '.htm', '.html' or '.cgi' as a page. A lot of sites will
probably define other extensions, such as '.phtml', '.php3' and '.pl'
as pages as well. Some people consider this number as the number of
'pure' hits... I'm not sure if I totally agree with that viewpoint.
Some other programs (and people :) refer to this as 'Pageviews'.
Each request made to the server comes from a unique 'site', which can
be referenced by a name or ultimately, an IP address. The 'sites'
number shows how many unique IP addresses made requests to the server
during the reporting time period. This DOES NOT mean the number of
unique individual users (real people) that visited, which is impossible
to determine using just logs and the HTTP protocol (however, this
number might be about as close as you will get).
Whenever a request is made to the server from a given IP address
(site), the amount of time since a previous request by the address
is calculated (if any). If the time difference is greater than a
pre-configured 'visit timeout' value (or has never made a request before),
it is considered a 'new visit', and this total is incremented (both
for the site, and the IP address). The default timeout value is 30
minutes (can be changed), so if a user visits your site at 1:00 in
the afternoon, and then returns at 3:00, two visits would be registered.
Note: in the 'Top Sites' table, the visits total should be discounted
on 'Grouped' records, and thought of as the "Minimum number of visits"
that came from that grouping instead. Note: Visits only occur on
PageType requests, that is, for any request whose URL is one of the
'page' types defined with the PageType option. Due to the limitation
of the HTTP protocol, log rotations and other factors, this number
should not be taken as absolutely accurate, rather, it should be
considered a pretty close "guess".
The KBytes (kilobytes) value shows the amount of data, in KB, that
was sent out by the server during the specified reporting period. This
value is generated directly from the log file, so it is up to the
web server to produce accurate numbers in the logs (some web servers
do stupid things when it comes to reporting the number of bytes). In
general, this should be a fairly accurate representation of the amount
of outgoing traffic the server had, regardless of the web servers
Note: A kilobyte is 1024 bytes, not 1000 :)
Top Entry and Exit Pages
The Top Entry and Exit tables give a rough estimate of what URL's
are used to enter your site, and what the last pages viewed are.
Because of limitations in the HTTP protocol, log rotations, etc...
this number should be considered a good "rough guess" of the actual
numbers, however will give a good indication of the overall trend in
where users come into, and exit, your site.
Notes on FTP Log Files
The Webalizer now supports ftp logs produced by wu-ftpd and others, as
a standard 'xferlog'. To process an ftp log, you must either use the
-Ff command line option or have "LogType ftp" in your configuration file.
Support for additional formats may be forthcoming, however a future
version of the Webalizer is in the works that will allow user defined
log formats, so this will become a non-issue. It is recommended that
you create a separate configuration file for ftp analysis, since the
values used for your web server will most likely not be suited for ftp
log analysis (ie: page types, hostname, etc.. should be different).
Because of the difference in web and ftp logs, there are a few limitations:
- Because there is no concept of a 'response code' in ftp world, response
codes are restricted to either 200 (OK) or 206 (Partial Content), based
on the completion status found in xferlog (for wu-ftpd, 'i'=incomplete
and will generate a 206, 'c'=complete and will generate a 200). If your
ftp server doesn't supply the completion status, all requests will be
assigned a response code of 200. This allows the usage graph to display
all transfer requests (hits), and how many of those completed in success
(files - ie: 200 response codes).
- Page totals won't accurately reflect reality, since there isn't really
the concept of a 'page' in regards to ftp services. I have found that
setting the PageType value to "README", "FIRST", etc... seems to work
fairly well however, and will give a pretty good indication of how
many 'non-binary' files were requested. Of course, the content of your
ftp site will be different, so your results may vary.
- Visit totals also won't accurately reflect reality, since visits are
triggered on PageType requests (see above). What you usually wind up
with is visits=sites in most cases.
- Entry/Exit pages will not be calculated for ftp logs.
- For obvious reasons, referrers and user agents are not supported.
- You _cannot_ analyze both web and ftp logs at the same time.. they must
be done separately in different runs.
Notes on Referrers
Referrers are weird critters... They take many shapes and forms, which makes
it much harder to analyze than a typical URL, which at least has some
standardization. What is contained in the referrer field of your log
files varies depending on many factors, such as what site did the referral,
what type of system it comes from and how the actual referral was generated.
Why is this? Well, because a user can get to your site in many ways... They
may have your site bookmarked in their browser, they may simply type your
sites URL field in their browser, they could have clicked on a link on some
remote web page or they may have found your site from one of the many search
engines and site indexes found on the web. The Webalizer attempts to deal
with all this variation in an intelligent way by doing certain things to
the referrer string which makes it easier to analyze. Of course, if your
web server doesn't provide referrer information, you probably don't really
care and are asking yourself why you are reading this section...
Most referrer's will take the form of "http://somesite.com/somepage.html",
which is what you will get if the user clicks on a link somewhere on the
web in order to get to your site. Some will be a variation of this, and
look something like "file:/some/such/sillyname", which is a reference from
a HTML document on the users local machine. Several variations of this can
be used, depending on what type of system the user has, if he/she is on
a local network, the type of network, etc... To complicate things even
more, dynamic HTML documents and HTML documents that are generated by
CGI scripts or external programs produce lots of extra information which
is tacked on to the end of the referrer string in an almost infinite number
of ways. If the user just typed your URL into their browser or clicked on
a bookmark, there won't be any information in the referrer field and will
take the form "-".
In order to handle all these variations, The Webalizer parses the referrer
field in a certain way. First, if the referrer string begins with "http",
it assumes it is a normal referral and converts the "http://" and following
hostname to lowercase in order to simplify hiding if desired. For example,
the referrer "HTTP://WWW.MyHost.Com/This/Is/A/HTML/Document.html" will become
"http://www.myhost.com/This/Is/A/HTML/Document.html". Notice that only the
"http://" and hostname are converted to lower case... The rest of the
referrer field is left alone. This follows standard convention, as the
actual method (HTTP) and hostname are always case insensitive, while the
document name portion is case sensitive.
Referrers that came from search engines, dynamic HTML documents, CGI
scripts and other external programs usually tack on additional information
that it used to create the page. A common example of this can be found
in referrals that come from search engines and site indexes common on the
web. Sometimes, these referrers URL's can be several hundred characters
long and include all the information that the user typed in to search for
your site. The Webalizer deals with this type of referrer by stripping
off all the query information, which starts with a question mark '?'.
The Referrer "http://search.yahoo.com/search?p=usa%26global%26link" will
be converted to just "http://search.yahoo.com/search".
When a user comes to your site by using one of their bookmarks or by
typing in your URL directly into their browser, the referrer field is
blank, and looks like "-". Most sites will get more of these referrals
than any other type. The Webalizer converts this type of referral into
the string "- (Direct Request)". This is done in order to make it easier
to hide via a command line option or configuration file option. This is
because the character "-" is a valid character elsewhere in a referrer
field, and if not turned into something unique, could not be hidden without
possibly hiding other referrers that shouldn't be.
Notes on Visits/Entry/Exit Figures
The majority of data analyzed and reported on by The Webalizer is
as accurate and correct as possible based on the input log file.
However, due to the limitation of the HTTP protocol, the use of
firewalls, proxy servers, multi-user systems, the rotation of your
log files, and a myriad of other conditions, some of these numbers
cannot, without absolute accuracy, be calculated. In particular,
Visits, Entry Pages and Exit Pages are suspect to random errors
due to the above and other conditions. The reason for this is
twofold, 1) Log files are finite in size and time interval, and
2) There is no way to distinguish multiple individual users apart
given only an IP address. Because log files are finite, they have
a beginning and ending, which can be represented as a fixed time
period. There is no way of knowing what happened previous to this
time period, nor is it possible to predict future events based on
it. Also, because it is impossible to distinguish individual users
apart, multiple users that have the same IP address all appear to
be a single user, and are treated as such. This is most common where
corporate users sit behind a proxy/firewall to the outside world,
and all requests appear to come from the same location (the address
of the proxy/firewall itself). Dynamic IP assignment (used with
dial-up internet accounts) also present a problem, since the same
user will appear as to come from multiple places.
For example, suppose two users visit your server from XYZ company,
which has their network connected to the Internet by a proxy server
'fw.xyz.com'. All requests from the network look as though they
originated from 'fw.xyz.com', even though they were really initiated
from two separate users on different PC's. The Webalizer would
see these requests as from the same location, and would record only
1 visit, when in reality, there were two. Because entry and exit
pages are calculated in conjunction with visits, this situation
would also only record 1 entry and 1 exit page, when in reality,
there should be 2.
As another example, say a single user at XYZ company is surfing
around your website.. They arrive at 11:52pm the last day of
the month, and continue surfing until 12:30am, which is now a
new day (in a new month). Since a common practice is to rotate
(save then clear) the server logs at the end of the month, you
now have the users visit logged in two different files (current
and previous months). Because of this (and the fact that the
Webalizer clears history between months), the first page the
user requests after midnight will be counted as an entry page.
This is unavoidable, since it is the first request seen by that
particular IP address in the new month.
For the most part, the numbers shown for visits, entry and exit
pages are pretty good 'guesses', even though they may not be 100%
accurate. They do provide a good indication of overall trends,
and shouldn't be that far off from the real numbers to count much.
You should probably consider them as the 'minimum' amount possible,
since the actual (real) values should always be equal or greater
in all cases.