Date: Fri, 16 Apr 1999 10:45:04 -0400 (EDT)
From: Ryan Bloom <rbb@raleigh.ibm.com>
To: new-httpd@apache.org
Subject: Re: cgi performance in Apache-apr.
In-Reply-To: <14103.19410.945836.410627@zap.ne.mediaone.net>
Message-ID: <Pine.A41.4.05.9904161033280.16448-100000@webdev7.raleigh.ibm.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: new-httpd-owner@apache.org
Precedence: bulk
Reply-To: new-httpd@apache.org


Not a problem, I've been meaning to update the list.  I am writing a
module that creates a cgi spawning process.  Basically, when the server
starts, I fork a new process which has a pipe to each child server.  When
the children get a cgi request, they pipe their unique thread number
through the pipe, and open three named pipes specified by their thread_id.
The cgi server then forks a new process, which also opens the same three
pipes, and waits for data from the child server.  The child packages up
everything the cgi server will need to execute the cgi.  (filename,
arguments, environment, variables, suExec data, etc.), and sends it to the
cgi process.  The cgi process spilts everything out of its packaged
state, and executes the program properly.

The module was working a few weeks ago, but I never got a chance to
optimize it, and it wasn't showing a big performance improvement, because
I was writing data to the pipes in very small chunks, for debuggin
purposes.  Since then, I have been doing apr work.  But, I am devoting
today to the module, and I hope to have it working better soon.  When it
is ready for general consumption, I'll post it to new-httpd for everybody
to see and critique.

I don't think the per-dir env stuff is affecting the module at all, but I
just started looking at it again, so we'll see.

Ryan


On Fri, 16 Apr 1999, Ben Hyde wrote:

> 
> Ryan Bloom wrote:
>  > We did some very informal performance numbers for cgi's in the hybrid
>  > server on AIX, and we have found a problem.  The performance for AIX went
>  > downhill quickly when we tried running CGI scripts.  The reason for this,
>  > is that fork on AIX creates the full address space for the process.  This
>  > means, all of the threads are created ...
> 
> Excuse my curiosity but how has this situation evolved?  The per dir. 
> environment stuff came along afterware right?
> 
>  - ben
> 

_______________________________________________________________________
Ryan Bloom		rbb@raleigh.ibm.com
4205 S Miami Blvd	
RTP, NC 27709		It's a beautiful sight to see good dancers 
			doing simple steps.  It's a painful sight to
			see beginners doing complicated patterns.