Re: measuring processing time

Bill Snapper (snapper@OpenMarket.com)
Tue, 03 Jun 1997 06:24:37 -0400

Message-Id: <3393F0E5.2754@openmarket.com>
Date: Tue, 03 Jun 1997 06:24:37 -0400
From: Bill Snapper <snapper@OpenMarket.com>
To: Raul Gonzalez Barron <barron@EESUN2.tamu.edu>
Subject: Re: measuring processing time

The performance depends on several issues.

1) What web server are you running on?

2) Are you using the cgi-fcgi "shim" application to talk to FastCGI
   or are you using a native plugin?

The second question is the one I'm really interested in.  If you are
using the "shim" then you will not get the benefit of speed with 
regard to the process loading time as you still have to launch a process
each time.  You will however be able to take advantage of process
persistence regardless of which mechanism.

If you're using a native plugin to talk to FastCGI, the connect time
could be a factor if the server is not maintaining a persistent 
connection between the web server and the application.  Persistent
connections are part of the FastCGI design but not exploited in all
cases for several reasons.  The main reason is that when you have a 
server architecture which uses multiple listener processes for http
requests, you typically use a new connection per request if you're
sharing a pool of FastCGI applications amongst all the listeners.  This
is not mandatory but based on how a lot of FastCGI applications are 
coded today and most server support, this is what has been used.

Hope this information is useful.


- Bill -

-- 
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Bill Snapper           | Open Market, Inc.   | My opinions are my own,
Software Engineer      | 245 First St.       | and not necessarily
snapper@openmarket.com | Cambridge, MA 02142 | those of my employer, 
1-617-949-7365         |                     | Open Market Inc.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++