Memory Footprint Matters June 28, 2010
What does a smaller memory footprint mean for you? Although it does not directly translate to better performance, it allows your web site handle larger traffic. By how much? Let's look at some metrics, numbers and their meanings.
When is memory used?
The life cycle of a web page is rather short. When a request is received at the server, a chunk of memory is allocated to generate the content. The allocated memory is then discarded as the rendered content is sent back to the web browser. Depending on the nature of the web page, the rendering process usually take from a few milliseconds to a couple of seconds. If multiple requests are being handled at the same time, the server needs access to more free memory.
How much memory do my web pages need?
Different scripting languages have various ways to determine how much memory a script, or specific objects inside the script consumes. In this article we focus on PHP pages. With proper setup, the memory usage can be easily obtained by adding the following line to the end of a script (usually where the page footer is displayed):
Now let's convert this number to kilobytes (KB) by dividing it by 1024. Then divide this number again by 256. For example
The number of 256K segments required to run a script gives us a convenient way to remember and compare the memory requirement in various web applications. Some open source applications are measured in Windows/Apache2/PHP5:
If you would like to add more applications to the above list, please contact us.
It's not just about sizes
An application that has more features has more complex implementation. This doesn't mean, however, that every page in this application should bear the same overhead. In the above example, the login page of OpenX alone takes 87 segments! In order to make a page only consume what it really needs, the program has to be properly modularized. So reducing memory footprint is not just about sizes, it's about thoughtful architecture.
Can your server handle the traffic?
Assuming all your server memory can be used to generate web pages, how many concurrent requests can you process? For a page that takes 10 segments, a 512M server can handle 205 requests simultaneously. If the page uses only 1 segment, 2048 requests can be process at once, assuming no other bottlenecks. The difference is day and night.
Memory usage isn't everything
As pointed out earlier, smaller memory footprint doesn't mean better performance. In the context of interpreted web language such as PHP, however, the script execution is often shorter (better). In addition to memory usage, the rendered content size, number of network connections, complexity of SQL queries and client-side computation requirement also deserve great attention. We'll cover these topics in future articles.
When is memory used?
The life cycle of a web page is rather short. When a request is received at the server, a chunk of memory is allocated to generate the content. The allocated memory is then discarded as the rendered content is sent back to the web browser. Depending on the nature of the web page, the rendering process usually take from a few milliseconds to a couple of seconds. If multiple requests are being handled at the same time, the server needs access to more free memory.
How much memory do my web pages need?
Different scripting languages have various ways to determine how much memory a script, or specific objects inside the script consumes. In this article we focus on PHP pages. With proper setup, the memory usage can be easily obtained by adding the following line to the end of a script (usually where the page footer is displayed):
<?php
echo memory_get_usage(true);
?>
And the number you see is the amount of memory, in bytes, the script requires to run.echo memory_get_usage(true);
?>
Now let's convert this number to kilobytes (KB) by dividing it by 1024. Then divide this number again by 256. For example
1310720 / 1024 / 256 = 5
By default, PHP allocates memory in increments of 256K. Each increment is a "segment". This value can be set via the ZEND_MM_SEG_SIZE variable.
The number of 256K segments required to run a script gives us a convenient way to remember and compare the memory requirement in various web applications. Some open source applications are measured in Windows/Apache2/PHP5:
| Gyroscope (sample app) | x1 | |
| phpMyAdmin | x7 | |
| Code Ignitor | x19 | |
| Joomla | x55 | |
| Word Press | x66 | |
| OpenX | x87 |
If you would like to add more applications to the above list, please contact us.
It's not just about sizes
An application that has more features has more complex implementation. This doesn't mean, however, that every page in this application should bear the same overhead. In the above example, the login page of OpenX alone takes 87 segments! In order to make a page only consume what it really needs, the program has to be properly modularized. So reducing memory footprint is not just about sizes, it's about thoughtful architecture.
Can your server handle the traffic?
Assuming all your server memory can be used to generate web pages, how many concurrent requests can you process? For a page that takes 10 segments, a 512M server can handle 205 requests simultaneously. If the page uses only 1 segment, 2048 requests can be process at once, assuming no other bottlenecks. The difference is day and night.
Memory usage isn't everything
As pointed out earlier, smaller memory footprint doesn't mean better performance. In the context of interpreted web language such as PHP, however, the script execution is often shorter (better). In addition to memory usage, the rendered content size, number of network connections, complexity of SQL queries and client-side computation requirement also deserve great attention. We'll cover these topics in future articles.
Antradar 2024 ©
Antradar Software Inc.
Antradar Software Inc.
1-888-404-1337
http://www.antradar.com
Toronto,
ON M8V1J5
Canada