AWStats logfile analyzer 6.95 Documentation

Benchmarks

AWStats update process must be ran frequently, so it's important to know what is AWStats speed to choose an optimum delay between each update process according to AWStats speed and the refresh rate you need to have.
AWStats speed depends on AWStats version and options/setup you use in configuration file.

This is benchmark results with AWStats version 6.0 and a common configuration:

HARDWARE: Athlon 1 GHz / 256MB
SOFTWARE: Windows 2000 / Perl 5.8 (Cygwin Perl)
CONFIG OPTIONS: Default values were used: LogFormat=1, DNSLookup=0, URLWithQuery=0, URLReferrerWithQuery=0, URLWithAnchor=0, No plugins
AVERAGE SPEED: 5200 lines by seconds
Other times for different kind of web sites sizes are shown later in this page...

This is other important information to know:

- A log file size is about 150 (NCSA common/CLF log files) to 320 times (NCSA extended/XLF/ELF log files) its number of lines,
- 1,000 visits = 8,000 pages (with 8 pages/visits) = 64,000 lines (with 8 hits/page) = 20 MB file => 15 seconds (Athlon 1GHz, Standard Perl 5.8)
- History files (AWStats database, resuming the log analysis) has the following size (one file a month) : 15000+90*x+100*y bytes (where x is number of unique visitors a month and y is number of different pages on web sites). If you use option BuildHistoryFormat=xml, you must multiplie this value by 3.

WARNING ! All those data are average values for a common public site with default configuration. Calculation rule can be seriously changed according to web server or AWStats configuration and web site content.

Don't forget that benchmarks of log analyzers are made without reverse DNS lookup because DNS lookup is so slow (depending on Internet network and your system), that if enabled in AWStats configuration file, it would take more than 99% of the time of a log analysis ! Take a look at the following chart to:
- Get more real ideas on benchmarks results
- Get more information and advice on a good setup for your site.

This is examples of frequency/parameters you should use to have a good use of AWStats:

Your Web site traffic	Perl distrib	Values for parameters			Recommended update frequency (Rotate log delay)	Memory required**	Update process duration***
Your Web site traffic	Perl distrib	DNSLookup*	URLWithQuery	URLReferrerWithQuery	Recommended update frequency (Rotate log delay)	Memory required**	Update process duration***
0 - 1,000 visits/month	Your choice	0 (or 2)	0 or 1	0 or 1	Once a day Log files are 0-1 MB 2000 lines to process	4 MB free	1s
0 - 1,000 visits/month	Your choice	1	0 or 1	0 or 1	Once a day Log files are 0-1 MB 2000 lines to process	4 MB free	2mn
1,000 - 10,000 visits/month	Your choice	0 (or 2)	0 or 1	0 or 1	Once a day Log files are 1-10 MB 2000-20000 lines	4-8 MB free	1s-10s
1,000 - 10,000 visits/month	Your choice	1	0 or 1	0 or 1	Once a day Log files are 1-10 MB 2000-20000 lines	4-8 MB free	2mn-10mn
10,000 - 100,000 visits/month	Your choice	0 (or 2)	0 or 1	0 or 1	One a day Log Files are 10-100 MB 20000-200000 lines	8-32 MB free	10s-120s
10,000 - 100,000 visits/month	Your choice	1	0 or 1	0 or 1	Once a day Log files are 10-100 MB 20000-200000 lines	8-32 MB free	10mn-50mn
100,000 - 500,000 visits/month	Your choice	0 (or 2)	0	0 or 1	Every 6 hours Log Files are 24-96 MB 50000-300000 lines	16-92 MB free	30s-3mn
100,000 - 500,000 visits/month	Your choice	1	0	0 or 1	Every 6 hours Log Files are 24-96 MB 50000-300000 lines	16-92 MB free	15mn-60mn
500,000 - 2,000,000 visits/month	See next section on pb with ActiveState	0 (or 2)	0	0	Every 6 hours Log Files are 96-384 MB 300000-1200000 lines	64-256 MB free	3mn-12mn
2,000,000 - 4,000,000 visits/month	See next section on pb with ActiveState	0 (or 2)	0	0	Every 6 hours Log Files are 384-768 MB 1200000-2400000 lines	256-512 MB free	12mn-24mn
+4,000,000 visits/month	See next section on pb with ActiveState	AWStats is a good choice for such web sites only if you use a dedicated server with a large amount of memory. Try a tool with less features but faster like "row counter log analyzers" like Analog if not.

* You should set DNSLookup parameter to 0 (or 2) if
- reverse DNS lookup is already done in your log file,
- or if your web site has more than 250,000 visits a month.
Note: Country report can works without reverse DNS lookup if plugin 'geoip' is enabled (faster and more accurate than reverse DNS lookup).
** This is free memory required for update process (in MB), this is not hardware memory installed !
Warning: If you use the URLWithQuery or URLReferrerWithQuery option, or forget to complete correctly URLQuerySeparators for some sites, this value can be dramatically increased.
*** Duration with DNSLookup set to 1 is very long because of DNS lookup whatever is speed of your computer. Duration with DNSLookup set to 0 (or 2) is with Athlon 1GHz/256MB, Cygwin Perl 5.8 and LogFormat=1.

SOME IMPORTANT ADVICES FOR A GOOD USE OF AWSTATS:

- Check that DNSLookup is disabled in AWStats (DNSLookup should not be set to 1). If you need a 'Country' report, you should prefer using the 'geoip' plugin using Maxmind database instead of a DNS lookup. Those plugins allow you to have more accurate results, faster with no network queries (With DNSLookup enabled, log analyze speed is decreased by 40 to 100 times, so use it only if required). Note that without DNS lookup and without the 'geoip' plugins, 'Country' report might work but results will be less accurate than the 'geoip' usage and it works only if hosts addresses in your log file are already resolved (need to setup your web server to do so, your web server will be slowed).

- Use carefully parameters URLWithQuery, URLReferrerWithQuery and URLWithAnchor (Let them set to 0 if you don't know what they means) and check your web site URLs' syntax to know if you don't need to complete the parameter URLQuerySeparators. If you really need to use URLWithQuery=1, check that URLWithQueryWithOnlyFollowingParameters or URLWithQueryWithoutFollowingParameters are set properly.

- Use last Perl version (For example Perl 5.8 is 5% faster than 5.6) and, more important, for large log files, use standard Perl distribution instead of ActiveState. This is because ActiveState 5.006 (and may be also other versions), has very important memory hole problem making speed of analysis slower and slower reaching 0 lines/seconds and using all your memory. You can see the decrease by adding the -showsteps option on command line. The speed should be constant to value given in top of this page, even for several Gigabytes log files !

- Rotate your log (See FAQ-SET500) and launch AWStats more often (from crontab or a scheduler, See FAQ-SET550). The more often you launch AWStats, the less AWStats has new lines in log to process. This can also solve the ActiveState memory problem (see next advice).

- Be sure that your HostAliases parameter list is complete.

- Use last AWStats version (For example AWStats 6.0 is 15% faster than 5.9).

- For geeks users, you can also recompile your Perl with differents options. For example, it seems that the "use64bitint=define usemymalloc=y" can increase speed by 10%.