Boosting The Performance of your

PHP Applications
PHP International Conference 2002

November 4, 2002. Frankfurt, Germany

Sterling Hughes & Derick Rethans <br />

<sterling@bumblebury.com /
d.rethans@jdimedia.nl>
Slide 1/40 November 4, 2002
Introduction

o PHP optimization
o Server side caching
o Database optimization
o Bandwidth optimization
o External utilities

-2-
Slide 2/40 November 4, 2002
Reducing Output
Less output is good because...
o Saves server bandwidth
o Decreases server CPU and Memory<br>usage (well, kinda)
o Has the same effect on the client side
o On large sites, bandwidth is most<br>often the most costly bottleneck

-3-
Slide 3/40 November 4, 2002
Optimize

o Use preprocessors to eliminate<br>whitespace

o If you save 20 bytes per file, and have 1,000,000 hits a day, you've just saved 20,000,000 extra
bytes
o Send content "when ready", and use progressive images

-4-
Slide 4/40 November 4, 2002
HTML Optimization
o Remove All Comments!
o Meta tags don't work!
o Use parent HTML tags to specify child values, when all children have the same value
o Fully qualify links (ie, personal/images/<br>not personal/images)

-5-
Slide 5/40 November 4, 2002
CSS Optimization

o Use CSS instead of incessant <font> tag usage

o All CSS should be contained in an external file that can be cached.
o Use short names for CSS attributes
o Override global HTML tag attributes using CSS properties
o Use CSS to control table definitions with CSS v2...

<STYLE TYPE="text/css">
<--
TABLE {table-layout:fixed; width: 100%}
#fixed COL {width: 50%}
-->
</STYLE>

<TABLE>
<COL><COL>
<TR><TD>Cell 1</TD><TD>Cell 2</TD></TR>
<TR><TD>Cell 3</TD><TD>Cell 3</TD></TR>
</TABLE>

-6-
Slide 6/40 November 4, 2002
HTTP Optimization
URL Shortening
By using Apache's <i>mod_rewrite</i> module, you can greatly shorten your HTML files by
making common locations into one/two letter aliases.

Apache Mod_Rewrite Entry

RewriteMap abbr dbm:/etc/apache/abbr

RewriteRule ^/r/([^/])/?(.) ${abbr:$1}$2
[redirect=permanent,last]

/etc/apache/abbr

b bytes/
d dogs/
g graphics/

gt graphics/template/

HTTP Caching
Use HTTP headers to have finer grained control over browser caching, these headers include:

Headers
Last-Modified: Mon, 22 Jul 2002 15:50:00 GMT
Expires: Mon, 22 Jul 2002 15:52:25 GMT
Cache-Control: must-revalidate, max-age=15, s-maxage=0, private

-7-
Slide 7/40 November 4, 2002
Compress
HTTP supports native compression by using Lempel-Ziv encoding (LZ77). Its the same
encoding mechanism used by the popular unix tool <code>gzip</code>.

Compression Advantages
o Greatly reduces bandwidth
o Cross browser compliant
o Compression can cache as well

Compression Disadvantages
o CPU intensive on the Server
o CPU intensive on the Client

-8-
Slide 8/40 November 4, 2002
PHP Compression

PHP supports output compression via PHP's ouput buffering features and the zlib extension.

The simplest way is to use the <code>ob_start ('ob_gzhandler')</code> or use the configuration
options <code>zlib.output_compression</code> and <code>zlib.compression_level</code>

HTTP_Compress
Alternatively, you can use PEAR's excellent HTTP_Compress class that will handle everything
related to HTTP_Compression for you...
<?php
require_once 'HTTP/Compress.php';

HTTP_Compress::start ();
print "Some data\n";
print "Some more data\n";
print "Even more data\n";
HTTP_Compress::output ();
? >

-9-
Slide 9/40 November 4, 2002
mod_gzip

Mod_gzip is a superb apache module which handles gzip compression and data caching. It
offers signifigant increases in compression performance to the homebrewed PHP solutions.

mod_gzip features
o Compression statistics
o The ability to chose compression via MIME type<br> and Filename
o You can exclude compression for certain browsers
o Can cache compression results

- 10 -
Slide 10/40 November 4, 2002
Server side caching

o Caching dynamic content

o Pregenerating content
o Creating pages on demand

- 11 -
Slide 11/40 November 4, 2002
Do not use PHP!

PHP...
&nbsp;&nbsp;&nbsp;...leads to slower pages.
Slower pages...
&nbsp;&nbsp;&nbsp;...lead to fewer users.
Fewer users...
&nbsp;&nbsp;&nbsp;...lead to less income.

- 12 -
Slide 12/40 November 4, 2002
Static rulez!

How dynamic is your content?

- 13 -
Slide 13/40 November 4, 2002
Do use PHP!

but
o Use as much static content as possible
o Cache dynamic generated content
o Use as few database queries as possible

- 14 -
Slide 14/40 November 4, 2002
PEAR::Cache_Lite

o Simple caching
o Caches pages, blocks, functions

- 15 -
Slide 15/40 November 4, 2002
PEAR::Cache_Lite - Example

<?php
/ Include the class /
require_once('Cache/Lite.php');

/ Set a key for this cache item /

$id = 'newsitem1';

/ Set a few options /

$options = array(
'cacheDir' => '/var/cache/news/',
'lifeTime' => 3600
);

/ Create a Cache_Lite object /

$Cache_Lite = new Cache_Lite($options);

/ Test if there is a valid cache-entry for this key /

if ($data = $Cache_Lite->get($id)) {
/ Cache hit! /
} else {
/ Cache miss! /
ob_start();
/ Generate content /
$data = ob_get_contents();
$Cache_Lite->save($data);
}
? >

- 16 -
Slide 16/40 November 4, 2002
Caching: Pros and Cons
Pros:
o Less use of external resources
Cons:
o PHP is still parsing the page
o Cache expiry time
o Stale content

- 17 -
Slide 17/40 November 4, 2002
Pregenerated content

o Only regenerate page when needed

o Different elements

- 18 -
Slide 18/40 November 4, 2002
Pregenerating: Pros and Cons
Pros:
o No stale content
o PHP is only used for regeneration
Cons:
o Costs a lot of resources
o Hard to figure out what to regenerate

- 19 -
Slide 19/40 November 4, 2002
Creating pages on demand

o Only regenerate page when requested

o Most pages are never requested
o Use the 404-trick

- 20 -
Slide 20/40 November 4, 2002
On demand - example
Facts:

Url: http://yoursite.com/news/000001.html
Path: /home/httpd/news/

Set up a 404 error handler:

<Directory /home/httpd/news>
ErrorDocument 404 /generate-news.php
</Directory>

generate-news.php:
<?php
$id = basename($_SERVER['REDIRECT_URL'], '.html');

/ Generate the HTML /

include "/home/httpd/news/generator.php";
$html = generate_html($data);

/ Show the page /

echo $html;

/ Store the page for later use /

$fp = fopen(sprintf("/home/httpd/news/%06d.html", $id), "w");
fputs($fp, $html);
fclose($fp);
? >

- 21 -
Slide 21/40 November 4, 2002
On demand: Pros and Cons
Pros:
o No wasted resources on generating
o No stale content
Cons:
o None :-)

- 22 -
Slide 22/40 November 4, 2002
External Tools

o Bytecode caches
o Content caches
o SRM: Script Running Machine
o lingerd

- 23 -
Slide 23/40 November 4, 2002
Bytecode caches

o Cache compiled PHP scripts

o Serializing caches:
o APC (Alternative PHP Cache)
o 'SHM' caches:
o ionCube Accelerator
o Zend Cache

- 24 -
Slide 24/40 November 4, 2002
Cache Benchmark

- 25 -
Slide 25/40 November 4, 2002
Squid
o transparent caching
o caches HTTP and FTP

Some effort to use with PHP:

<?php
$tsstring = gmdate("D, d M Y H:i:s ", $timestamp)."GMT";

if ($_SERVER["HTTP_IF_MODIFIED_SINCE"] == $tsstring) {
/ The UA has the exact same page we have. /
header("HTTP/1.1 304 Not Modified");
exit();
} else {
header("Last-Modified: ".$tsstring);
}
? >

- 26 -
Slide 26/40 November 4, 2002
SRM: Script Running Machine

o Functions to work with persistent Objects

o Bridge between PHP Client and Objects
o Objects run in threads
o Manage persistent resources and data

- 27 -
Slide 27/40 November 4, 2002
lingerd

o Lingering conections
o Apache hands over the job
o Apache can do useful things
o Less server load

- 28 -
Slide 28/40 November 4, 2002
Performance
So You Want to get performance from your php applications?

- 29 -
Slide 29/40 November 4, 2002
Sockets
Performance Problems
o Makes you network dependent (TCP/IP)
o Communication protocol overheads (TCP/IP, especially)
Performance Tips
o When dealing with local sockets consider other<br> options like pipes or shared memory.
o Establishing a connection is the most costly <br> part of the TCP/IP layer. Think carefully about
re-using pre-existing connections.
o When transfering large chunks of data, consider using a compression/encryption method to lower
bandwidth

- 30 -
Slide 30/40 November 4, 2002
File I/O
File I/O Problems
o Disk access is slower than memory access (duh!)
o Simultaeneous write/write operations cause data corruption. Therefore locking must occur.
o Locking is slow
File I/O Tips
o Use the disk in concentrated bits, meaning do all<br>your writing together.
o Do Not Read Large Files Into Memory!
o Do Not Read Small Files From Disk!
o Employ caching to access frequently used data<br> stored in memory
o Try and put write operations where they will<br> cause the least loss in user experience
(remember,<br> web pages are chunked!)

- 31 -
Slide 31/40 November 4, 2002
CPU Usage
o Avoid Using frequent array accesses, remember, each array access is a hash lookup.
The Wrong Way
<?php
$ar['name'] = 'Sterling';

for ($i = 0; $i < 20; ++$i) {

print $ar['name'];
}
? >

o Shorten "required" executions, such as loops.

o Avoid rebuilding large data structures.
o Look for alternatives to regular expressions

- 32 -
Slide 32/40 November 4, 2002
Regular Expressions
Slow Regexps
Its often beneficial to replace regular expressions with simpler functions that have the same
effect.

With Regular Expression

<?php
if (preg_match ('/(.)@(.)/', $data, $matches)) {
print $matches[2];
}
? >

Without Regular Expressions

<?php
$pos = strrpos ('@', $data);
if ($pos !== false) {
print substr ($data, $pos);
}
? >

Fast Regexps
In other cases it may help to use regular expressions instead of custom parse routines in your PHP
code.

- 33 -
Slide 33/40 November 4, 2002
SQL Queries
SQL Problems
o Unused data is being pulled over a socket
o FULLTEXT indexes
o the LIKE clause
SQL Tips
o Study and Learn Database Optimization
o Failing that use the following tips...
o MySQL supports the 'LIMIT' clause which can be used<br> to lower the returned result set to a
specified number
o Never use LIKE or a FULLTEXT index to do complex<br> searches. Use keywords and an
alternative ranking system
o The only thing worse than using a LIKE clause, is building<br> your own LIKE clause in PHP.

- 34 -
Slide 34/40 November 4, 2002
Database optimization

o Queries are slow

o Recursive queries are slower
o Recursive queries on every page are slowest

- 35 -
Slide 35/40 November 4, 2002
Visitation model 1
Directory structure:

Stored as: <table class="db" border="1" align="center" cellspacing="0"> <tr>

<th>Id</th><th>Parent</th><th>Name</th> </tr> <tr> <td>3</td><td>1</td><td
class="left">Countries</td> </tr> <tr> <td>4</td><td>3</td><td class="left">Belgium</td>
</tr> <tr> <td>5</td><td>3</td><td class="left">Netherlands</td> </tr> <tr>
<td>6</td><td>3</td><td class="left">Germany</td> </tr> <tr> <td>7</td><td>5</td><td
class="left">Business and Economy</td> </tr> </table>

Retrieve list with:

<?php
function gimmeParent($id) {
global $dir;

$res = mysql_query("SELECT parent, name FROM directory WHERE id = $id");

if (mysql_num_rows($res) > 0) {
$dir[] = ($row = mysql_fetch_row($res));
gimmeParent($row['parent']);
}
}

gimmeParent(7);
? >

- 36 -
Slide 36/40 November 4, 2002
Visitation model 2
Tree structure:

Visitation:

- 37 -
Stored as: <table class="db" border="1" align="center" cellspacing="0"> <tr>
<th>Id</th><th>Parent</th><th>Name</th><th>Left</th><th>Right</th> </tr> <tr>
<td>3</td><td>1</td><td class="left">Countries</td><td>1</td><td>12</td> </tr> <tr>
<td>4</td><td>3</td><td class="left">Belgium</td><td>2</td><td>3</td> </tr> <tr>
<td>5</td><td>3</td><td class="left">Netherlands</td><td>4</td><td>9</td> </tr> <tr>
<td>6</td><td>3</td><td class="left">Germany</td><td>10</td><td>11</td> </tr> <tr>
<td>7</td><td>5</td><td class="left">Business</td><td>5</td><td>6</td> </tr> <tr>
<td>8</td><td>5</td><td class="left">Economy</td><td>7</td><td>8</td> </tr> </table>

- 38 -
Slide 37/40 November 4, 2002
Visitation model 3

Select the path:

Query:
SELECT * FROM directory
WHERE
left < 5 AND right > 6

- 39 -
Slide 38/40 November 4, 2002
Visitation model 4

Select all end points:

Query:
SELECT * FROM directory
WHERE
right - left = 1

- 40 -
Slide 39/40 November 4, 2002
Visitation model 5

Select a subtree:

Query:
SELECT * FROM directory
WHERE
left > 4 AND right < 9

- 41 -
Slide 40/40 November 4, 2002
Thank You!

These Slides: http://pres.derickrethans.nl

HTTP Cache Explanation: http://www.mnot.net/cache_docs/

HTML Tidy: http://www.w3.org/People/Raggett/tidy/
ionCube: http://www.ioncube.com
lingerd: http://www.iagora.com/about/software/lingerd
mod_gzip: http://www.remotecommunications.com/apache/mod_gzip/
PHP: http://www.php.net
squid: http://www.squid-cache.org
Webreference HTML Optimization: http://www.webreference.com/authoring/languages/html/optimize/
Zend: http://www.zend.com

- 42 -
 Index
Introduction ............................................................................................... 2
Reducing Output ....................................................................................... 3
Optimize .................................................................................................... 4
HTML Optimization ................................................................................. 5
CSS Optimization ...................................................................................... 6
HTTP Optimization ................................................................................... 7
Compress ................................................................................................... 8
PHP Compression ..................................................................................... 9
mod_gzip ................................................................................................... 10
Server side caching .................................................................................... 11
Do not use PHP! ........................................................................................ 12
Static rulez! ............................................................................................... 13
Do use PHP! .............................................................................................. 14
PEAR::Cache_Lite .................................................................................... 15
PEAR::Cache_Lite - Example .................................................................. 16
Caching: Pros and Cons ............................................................................ 17
Pregenerated content ................................................................................. 18
Pregenerating: Pros and Cons ................................................................... 19
Creating pages on demand ........................................................................ 20
On demand - example ............................................................................... 21
On demand: Pros and Cons ....................................................................... 22
External Tools ........................................................................................... 23
Bytecode caches ........................................................................................ 24
Cache Benchmark ..................................................................................... 25
Squid ......................................................................................................... 26
SRM: Script Running Machine ................................................................. 27
lingerd ........................................................................................................ 28
Performance .............................................................................................. 29
Sockets ...................................................................................................... 30
File I/O ...................................................................................................... 31
CPU Usage ................................................................................................ 32
Regular Expressions .................................................................................. 33
SQL Queries .............................................................................................. 34
Database optimization ............................................................................... 35
Visitation model 1 ..................................................................................... 36
Visitation model 2 ..................................................................................... 37
Visitation model 3 ..................................................................................... 39
Visitation model 4 ..................................................................................... 40
Visitation model 5 ..................................................................................... 41
Thank You! ............................................................................................... 42