Speeding Up Your Web Site Using mod_deflate Module for Apache

Published November 28th, 2008

Problem Statement

If your Web site serve a lot of textual data vs. image or video, you might be able to take advantage of mod_deflate and speed up your Web site without changing any PHP code or MySQL database schema. OK, sure it works for non-PHP/MySQL sites too. But hey, we are biased as we do PHP/MySQL for a living.
In this article, we will show you how to use mod_deflate to compress your textual contents to speed up delivery time, which results in bandwidth savings in addition to higher user sanctification.

Step 1: Did you install mod_deflate

Make sure that you have mod_deflate already installed as part of your Apache Web server. To check you can run /path/to/bin/httpd -l | grep mod_deflate and if you get mod_deflate.c back as a return string, you are in good shape. Otherwise, you will have to install the mod_deflate module for Apache.

Step 2: Creating a deflate.conf configuration file

Copy the following Apache directives in a file called deflate.conf and put it in your Apache conf directory.

<IfModule deflate_module>

<Location />
   SetOutputFilter DEFLATE
   SetEnvIfNoCase Request_URI \.(?:gif|jpe?g|png)$ no-gzip dont-vary
   SetEnvIfNoCase Request_URI \.(?:exe|t?gz|zip|bz2|sit|rar)$ no-gzip dont-vary
   SetEnvIfNoCase Request_URI \.pdf$ no-gzip dont-vary

   BrowserMatch ^Mozilla/4 gzip-only-text/html
   BrowserMatch ^Mozilla/4\.0[678] no-gzip
   BrowserMatch \bMSIE !no-gzip !gzip-only-text/html
</Location>

# Testing purposes (remove from production environment)
DeflateFilterNote Input inbytes
DeflateFilterNote Output outbytes
DeflateFilterNote Ratio percent
LogFormat '%h "%r" %{outbytes}n/%{inbytes}n (%{percent}n%%)' deflate
CustomLog logs/deflate.log deflate env=!no-gzip
</IfModule>

Here we have a Apache configuration specific to mod_deflate that only works if mod_deflate is already loaded by Apache. This configuration snippet tells Apache the following:

• Set the output filter for all files to be DEFLATE — which is the mod_deflate code that compresses contents passed on to the filter code
• Sets environment variable no-gzip for common image files such as .gif, .jpg, .jpeg, .png
• Sets environment variable no-gzip for common binary/compressed files such as .exe, .gz, .zip, .bz2, .sit and .rar
• Sets environment variable no-gzip for PDF file with .pdf extension
• Ignores older browsers that do nto support on-the-fly decompression on their end
• For testing purposes, we set a note called inbytes which stores the input (content) size in byte
• The outbytes note stores the output (contents) bytes and the percent note stores the compression ratio (output / input) * 100
• A custom log format allows us to create a custom log that records all compressed (!no-gzip) requests

Step 3: Testing your setup

Now using your favorite Web browser, request compressible pages of your site and monitor the deflate.log in the logs directory of your Apache installation. The deflate.log should show how good the compression ratio. The better the ratio, the faster you are able to send out compressed contents to the Web clients.

Why is compressed HTML pages better?

In the Web’s request and response model, the bottlenecks are in the following order:

1. Network latency/bandwidth/response
2. Disk I/O
3. Other System Calls

By compressing the HTML contents, which adds to CPU processing time on the server side but no additional disk I/O and also reduces network response time as the size of the content is less than the original. It adds decompression processing on client side. So the additional cost of compression/decompression is only CPU processing on server and client side. Since modern CPU is hardly the bottleneck, the benefit is: network response time reduction.

See Also
Apache 2.0 mod_deflate documentation

Category: Performance | Tags: