SpeedyCGI Benchmarks

If you would like to contribute benchmarks, please send email to sam@daemoninc.com.
Date: Sat, 10 Feb 2001 16:33:07 +0100
From: Marc Lehmann 
To: sam@daemoninc.com
Subject: speedycgi benchmark

Hi, I just saw this in the TODO:

TODO
       ยท   Need benchmarks of speedy vs mod_perl

I of course can't provide this, but I just benchmarked speedycgi against
PApp::Apache (or better: PApp::Apache vs. PApp::CGI with speedycgi).

PApp has a *tremendous* startup-penalty and is (or was ;) usually run
using it's own optimized mod_perl handler. Since this can be quite
difficult to set-up (due to a lot of bugs in apache regarding dynamic
loading and expat-lite which nobody uses but causes everybody grief ;) I
finally wrote PApp::CGI and tested it. A complicated page gave me these
numbers:

PApp::CGI (pure cgi, apache):   1.5 hits/s
PApp::CGI (using speedycgi):    21 hits/s
PApp::Apache (native mod_perl): 30 hits/s

(this is the sustained hit rate with 5 clients)

All in all this is *very* respectable. I won't test mod_speedy (obviously
if I can have mod_speed I can use mod_perl anyway), but speedycgi
certainly makes for a very viable aim (almost no setup-penalty, just as
fast in practise). BTW, PApp::Apache does almost nothing with regards to
set-up while PApp::CGI mostly emulated the Apache API, so there might be a
(small) bias involved.

Thanks a lot ;)

-- 
      -----==-                                             |
      ----==-- _                                           |
      ---==---(_)__  __ ____  __       Marc Lehmann      +--
      --==---/ / _ \/ // /\ \/ /       pcg@goof.com      |e|
      -=====/_/_//_/\_,_/ /_/\_\       XX11-RIPE         --+
    The choice of a GNU generation                       |
                                                         |