Re: WWW BENCHMARKS for Hello World (fwd)

---------- Forwarded message ---------- Date: Wed, 17 Mar 1999 18:12:56 -0500 (EST) From: Rasmus Lerdorf <rasmus@lerdorf.on.ca> To: Ken Williams <ken@forum.swarthmore.edu> Cc: Joshua Chamas <joshua@chamas.com>, modperl@apache.org, Matthew.Sergeant@eml.ericsson.se, perl-win32-users@listserv.ActiveState.com Subject: [PHP-DEV] Re: WWW BENCHMARKS for Hello World On Wed, 17 Mar 1999, Ken Williams wrote: > Joshua Chamas's message: > >Here's a round-up of some testing done: > > > >Environment Hits/s OS Chip Sessions > >----------- ------ -- ---- -------- > >iis/static html 330.01 WinNT PII 300 no > >apache/static html 74.08 WinNT PII 300 no > >iis/perlex cgi 64.87 WinNT PII 300 no > >iis/perl isapi raw-cgi 59.78 WinNT PII 300 no > >apache/chiliasp vbscript 56.93 WinNT PII 300 yes > >apache/mod_perl 50.72 WinNT PII 300 no > >apache/mod_perl/cgi (registry) 40.17 WinNT PII 300 no > >apache/mod_perl/asp 27.76 WinNT PII 300 no > >apache/mod_perl/embperl 27.11 WinNT PII 300 no > >iis/perl raw-cgi 16.35 WinNT PII 300 no > >apache/mod_perl/asp 15.75 WinNT PII 300 yes > >iis/asp perlscript 9.54 WinNT PII 300 both > >iis/asp vbscript 9.51 WinNT PII 300 both > >iis/perl cgi 3.82 WinNT PII 300 no > >iis/perl isasp cgi 2.96 WinNT PII 300 no > > It would be cool to see the results of this same test on Unix. > > Anyone? =) We are comparing apples and oranges here. And benchmarks like these tend to be rather bogus. But, just to toss out some more meaningless numbers here I ran a similar test on a Linux-2.2.3 box. Same cpu. PII-300. I used the Apache benchmarking program (ab) since the one Joshua used only runs on Windows. The lines I did would be: Environment Hits/s OS Chip ----------- ------ -- ---- apache/static html 612.56 Linux PII 300 apache/mod_php 375.87 Linux PII 300 The details... First, the static html case: > ab -t 30 -c 10 http://localhost/hello.html Server Software: Apache/1.3.5-dev Server Hostname: localhost Server Port: 80 Document Path: /hello.html Document Length: 12 bytes Concurrency Level: 10 Time taken for tests: 30.005 seconds Complete requests: 18380 Failed requests: 0 Total transferred: 5002080 bytes HTML transferred: 220680 bytes Requests per second: 612.56 Transfer rate: 166.71 kb/s received Connnection Times (ms) min avg max Connect: 0 6 68 Processing: 12 9 8 Total: 12 15 76 Second, the dynamic case. I followed the rule and did half dynamic, half-static since php is an html-embedded language. The hello.php3 file contained: <?echo "Hello "?>World >ab -t 30 -c 10 http://localhost/hello.php3 Server Software: Apache/1.3.5-dev Server Hostname: localhost Server Port: 80 Document Path: /hello.php3 Document Length: 12 bytes Concurrency Level: 10 Time taken for tests: 30.010 seconds Complete requests: 11280 Failed requests: 0 Total transferred: 1795110 bytes HTML transferred: 135480 bytes Requests per second: 375.87 Transfer rate: 59.82 kb/s received Connnection Times (ms) min avg max Connect: 0 11 70 Processing: 15 14 12 Total: 15 25 82 So, both the static and the dynamic were significantly faster than anything on Windows. This is a highly unscientific test though, and I know for a fact that IIS competes very well on static content. Perhaps the file was too small to really benefit from IIS's optimizations though. I don't know. I just tried to recreate the test to the best of my abilities on Unix. I did not fudge anything nor go out of my way to optimize anything. -Rasmus -- PHP Development Mailing List http://www.php.net/ To unsubscribe send an empty message to php-dev-unsubscribe@lists.php.net For help: php-dev-help@lists.php.net