Re: can FCGI_STDIN content length be increased when receivingthe STDIN data?

Yeah, I did some research and it looks like APR_BUCKET_BUFF_SIZE 8000 is
your limit.  Apache developers did this to align data reads and writes with
system page sizes to maximize performance.  I didn't find a config option
for it because it could hinder performance if you tune it wrong. 

 

In order to correctly increase it you will have to recompile Apache with the
value set to be an exact multiple of 4KB in order to not affect performance
of Apache.  The setting is located in srclib/apr-util/buckets/apr_buckets.h
under the Apache source code.

 

A link that talks about is at
http://www.mail-archive.com/dev-XeBeRKkkxMkyzMRdD/[email protected]/msg44168.html

 

Martin

  _____  

From: fastcgi-developers-bounces+chapmanm=pixia.com-xGejAJT2w6xVgU18Zptdi0EOCMrvLtNR@public.gmane.org
[mailto:fastcgi-developers-bounces+chapmanm=pixia.com-xGejAJT2w6xVgU18Zptdi0EOCMrvLtNR@public.gmane.org]
On Behalf Of Konstantin Ivanov
Sent: Friday, December 11, 2009 8:37 PM
To: fastcgi-developers-xGejAJT2w6xVgU18Zptdi0EOCMrvLtNR@public.gmane.org
Subject: [FASTCGI] can FCGI_STDIN content length be increased when
receivingthe STDIN data?

 

Hi all, I have the following situation which I am try to bypass in attempt
to increase the performance. I am using PHP (php-cgi) in FCGI mode on
Windows. I have a client to read from the fastcgi server written in C++. The
"issue" I am experience is the size of the STDIN content length. I have a
file of 65000 bytes (actually it is a print from php-info() function), but
when I read it from the FastCGI process the STDIN type header always comes
back with the maximum of 8184 bytes, which means that in order to read the
entire request and before receiving FCGI_END_REQUEST i have to loop and read
from the socket fd 8 times. I think that this is very inconvenient and I was
wondering what controls the STDIN maximum content length? You see, if one
would increase the Windows socket window size to say 128Kb, then the request
of 65000 bytes could be read with a single iteration to recv() function. Did
anyone experience the same behavior and if so, would you please give your
advice or point in the right direction. 

 

Thank you,

_______________________________________________
FastCGI-developers mailing list
FastCGI-developers-xGejAJT2w6xVgU18Zptdi0EOCMrvLtNR@public.gmane.org
http://mailman.pins.net/mailman/listinfo.cgi/fastcgi-developers