httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alvaro Martinez Echevarria <>
Subject server survey
Date Fri, 10 Apr 1998 05:12:07 GMT
Hi there.
I am doing a web server survey on my spare time using some
thousands of servers extracted from the logs of a couple of
quite big proxies. The main reason to lose my time doing so is
that I think there's some interesting data we can't know after
netcraft's surveys. For example, the 'real' traffic that each
family of servers handles (and not just the number of machines
using a member of each family). Just an advance from the february

Server       percentage      percentage	        comments
            of requests      of servers
               served          running
Apache          45.0%           52.5%
Netscape        21.0%           15.4%
Microsoft       14.9%           14.9%
AV               2.8%              0%         (Altavista Server, I guess)
O'Reilly         1.4%            2.5%         (WebSite and WebSitePro)
NCSA             1.3%            4.5%
Roxen            0.6%            0.4%
OMSW             0.6%            0.1%         (Open-Market-Secure-Webserver)
CERN             0.3%            1.3%
unspecified      7.3%            1.1%         (no Server: header is given)
other            4.8%            7.3%

The analysis has been done over 5.3 million requests made on
43000 servers and registered by the proxies I have used. The
results are obviously influded by the fact that the proxies
(where I got the source data from) are being used by spanish
users and hence about the 20% of the requests are made to '.es'
servers (which represent a 6.4% out of the total number of
servers analyzed); and here in Spain Apache is getting much
lower results than it should:

Partial results (only in '*.es')
Server       percentage      percentage	        comments
            of requests      of servers
               served          running
Apache          33.9%           32.3%
Netscape        40.4%           22.2%
Microsoft       18.2%           25.7%

Partial results (in '*' excepting '*.es')
Server       percentage      percentage	        comments
            of requests      of servers
               served          running
Apache          47.6%           53.9%
Netscape        16.3%           14.9%
Microsoft       14.1%           14.2%

Of course, before getting the numbers I removed from the data
the requests made to my servers, it wouldn't be fair (and well,
we are using Apache).

There's something more I have found out after the data,
although it should be taken with a bit of salt. In a number of
servers (around 1000) I couldn't obtain the server identification
in the first connection because of different kinds of errors
(mostly DNS, timeout and connection reset problems), but I did in
a second trial. The percentages change quite a lot in this

Requests that resulted in errors
Server       percentage      comments
             of servers
Apache          33.3%        
Netscape        11.5%        
Microsoft       32.5%        :-)

Could this mean that the average sloppy and/or misconfigured site
runs Microsoft more frequently than the rest? Are Microsoft
servers much more prone to errors than Apache and Netscape?

I hope you've found this interesting. When I have time I'll put
some more data in a web page, and probably I'll try to make a new
survey every month.


|   Alvaro Martínez Echevarría   |      LANDER SISTEMAS            |
|        |      Pº Castellana, 121         |
`--------------------------------|      28046 Madrid, SPAIN        |
                                 |      Tel: +34-1-5562883         |
                                 |      Fax: +34-1-5563001         |

View raw message