[ILUG] aolserver vs apache

[Donncha O Caoimh]

Just sent this to webdev, but I think it's probably worth posting here
too.

----------------------------------------------------------------
Quick benchmark using Apache Bench.
Ran it with -n 1000 -c 100 parameters.
Fetched t.php, a file containing the following:
<?php
phpinfo();
?>
Also fetched root document. AOLserver default page on one machine vs a
page with one gif on the other machine.

Software tested:
AOLserver 3.1, php 4.0.2 on a 350MHz PII with 128MB ram which is also
serving Goldmine to a LAN.
Apache 1.3.12, php 4.0.0 on a 500MHz PIII  with 512MB ram. Runs a small
program in the background feeding data to other servers which only takes
between 2% to 5% of CPU and very little memory.

Note: AOLserver was tuned slightly to allow max 500 connections with a
minimum of 300.
Apache has the following:
MinSpareServers 5
MaxSpareServers 10
StartServers 5
MaxClients 250



Results:
1. t.php
Apache
Requests per second:    17.30
Transfer rate:          559.50 kb/s received
Document Length:        32148 bytes

AOLserver
Requests per second:    9.85
Transfer rate:          371.04 kb/s received
Document Length:        37037 bytes

2. root document
Apache
Requests per second:    17.39
Transfer rate:          3.77 kb/s received
Document Length:        52 bytes 

AOLserver
Requests per second:    583.77
Transfer rate:          1108.58 kb/s received
Document Length:        1689 bytes

Note: Apache Bench did download the gif file from the Apache server. It
reported traffic transferred of 217000 bytes, while HTML transferred was
52000 bytes as expected.
AOLserver transferred 1899000 bytes in total, 1689000 bytes of HTML.

Conclusion:
In this simple test, and in the context of using AOLserver as a PHP
aware webserver, Apache beats it easily. The PHP documentation
(php-4.0.2/sapi/aolserver/README) states that the AOLserver port is not
ready for production yet, but it's already more stable than it was a few
months ago.
AOLserver cannot be beaten by Apache for serving static content though. 

Apache Bench works much better if you stress test the servers over a
longer period building up number of hits over several minutes instead of
suddenly bombarding the server with 100 requests. Maybe next time.

Result:
Best for PHP: Apache 1.3
Best for static content: AOLserver 3.1

Lies, damn lies, and benchmarks.. do your own!

Donncha.