The Fastest OpenSource WebServer? [on hold]

The Fastest OpenSource WebServer? [on hold]

How can I improve performance of these servers?
cherokee-project.com
nginx.org
open.litespeedtech.com
Server Configuration
VirtualBox CentOS 6.4 64bit
Xeon W3540 @3.2Ghz allocated 4 cpu threads
1500MB memory
20GB disk on 1TB Western Digital Black Caviar
PHP info and locations
hxxp header comparisons
curl -I hxxp://FastTest.com/phpinfo.php
Server: nginx centminmod
Date: Sat, 24 Aug 2013 03:33:05 GMT
Content-Type: text/html; charset=utf-8
Connection: keep-alive
Vary: Accept-Encoding
curl -I hxxp://FastTest.com:8088/phpinfo.php
hxxp/1.1 200 OK
Content-type: text/html
Date: Sat, 24 Aug 2013 03:33:05 GMT
Accept-Ranges: bytes
Server: LiteSpeed
curl -I hxxp://FastTest.com:81/phpinfo.php
hxxp/1.1 200 OK
Connection: close
Date: Sat, 24 Aug 2013 03:33:06 GMT
Server: Cherokee
Content-type: text/html
Setup Configuration
Nginx 1.4.2 phpinfo.php (php-fpm 5.4.18 + APC Cache 3.1.13) =
hxxp://FastTest.com/phpinfo.php
OpenLiteSpeed 1.2.4 phpinfo.php (php-fpm 5.4.18 + APC Cache 3.1.13) =
hxxp://FastTest.com:8088/phpinfo.php
Cherokee 1.2.103 phpinfo.php (php-fpm 5.4.18 + APC Cache 3.1.13) =
hxxp://FastTest.com:81/phpinfo.php
Siege benchmark v3.0.2 Results
Run Siege benchmark v3.0.2 with following parameters:
siege -q -b -c25 -r50 hxxp://FastTest.com/phpinfo.php
siege -q -b -c25 -r50 hxxp://FastTest.com:8088/phpinfo.php
siege -q -b -c25 -r50 hxxp://FastTest.com:81/phpinfo.php
Results
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
SIEGE BENCHMARK V3.0.2 | ELAPSED TIME | DATA TRANSFERED |
RESPONSE TIME | TRANSACTION RATE | THROUGHPUT | LONGEST TRANSCATION |
SHORTEST TRANSACTION
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
Nginx 1 worker_process | 2.92 | 14.65 | 0.06
| 428.08 | 5.02 | 0.11 | 0.01
Nginx 4 worker_process | 2.05 | 14.65 | 0.04
| 609.76 | 7.15 | 0.11 | 0.01
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
OpenLiteSpeed with 1 worker | 2.89 | 12.67 | 0.06
| 432.53 | 4.38 | 0.10 | 0.01
OpenLiteSpeed with 1 worker | 1.41 | 12.67 | 0.03
| 886.52 | 8.98 | 0.10 | 0.00
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
Cherokee allow PHP caching | 2.29 | 14.50 | 0.05
| 545.85 | 6.33 | 0.13 | 0.02
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
Cherokee forbid PHP caching | 5.00 | 14.50 | 0.06
| 249.80 | 2.91 | 4.72 | 0.01
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
| | |
| | |
|
Conclusions
Here are my findings:
Nginx also can be configured to cache its PHP requests using
fastcgi_cache. Example setup for Centmin Mod and Wordpress with
fastcgi_cache. So for proper comparison with default Cherokee PHP cache
enabled setup, you would need to compare with Nginx and php-fpm configured
with fastcgi_cache.
Litespeed Enterprise server can do the same via LiteSpeed Cache.
Unfortuantely, OpenLiteSpeed free open source version of LiteSpeed
Enterprise has it's LiteSpeed Cache support removed. But it looks like it
performs fairly well with non-cached PHP requests already from above
benchmarks. You would need Litespeed Enterprise if you want to use
LiteSpeed Cache and cache PHP requests.
Cherokee seems to be default have PHP caching enabled although they do
have a note that mentions it's experimental in nature and to use with
caution. They also mention that by default front line cache doesn't cache
results with cookies. So not sure if phpinfo.php PHP file is the best
choice to benchmark against versus PHP files and PHP apps which utilise
cookies extensively ? No idea why even with PHP cache disabled, that Siege
benchmarks would fluctuate between non-cached and cached like results ? If
any Cherokee users know, please let me know via comments