Skip to content

o2system/benchmarks

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

alt text

O2System Framework Benchmark with Other PHP Frameworks

This project attempts to measure minimum overhead (minimum bootstrap cost) of PHP frameworks in the real world.

So I think the minimum applications to benchmark should not include:

  • cost of template engine (HTML output)
  • cost of database manipulation
  • cost of debugging information

Components like Template engine or ORM/Database libraries are out of scope in this project.

Benchmarking Policy

This is master branch.

  • Install a framework according to the official documentation.
  • Use the default configuration.
    • Don't remove any components/configurations even if they are not used.
    • With minimum changes to run this benchmark.
  • Set environment production/Turn off debug mode.
  • Run optimization which you normally do in your production environment, like Composer's --optimize-autoloader.
  • Use controller or action class if a framework has the functionality.

Some people may think using default configuration is not fair. But I think a framework's default configuration is an assertion of what it is. Default configuration is a good starting point to know a framework. And I can't optimize all the frameworks. Some frameworks are optimized, some are not, it is not fair. So I don't remove any components/configurations.

But if you are interested in benchmarking with optimization (removing components/configurations which are not used), See optimize branch.

If you find something wrong with my code, please feel free to send Pull Requests. But please note optimizing only for "Hello World" is not acceptable. Building fastest "Hello World" application is not the goal in this project.

Results

Benchmarking Environment

  • Fedora 28 64bit (Droptlet; Digital Ocean)
    • PHP 7.2
      • Zend OPcache v7.2.13
    • Apache/2.4.34 (Fedora)

Hello World Benchmark

These are my benchmarks, not yours. I encourage you to run on your (production equivalent) environments.

(2017/02/14)

Benchmark Results Graph

framework requests per second relative peak memory relative
fatfree-3.5 811.68 10.9 0.46 inf
slim-2.6 603.73 8.1 0.52 inf
ci-4.0-dev 1,030.29 13.8 0.00 nan
slim-3.0 519.23 7.0 0.00 nan
bear-1.0 425.33 5.7 0.65 inf
lumen-5.7 228.51 3.1 0.00 nan
ze-1.0 462.78 6.2 0.00 nan
yii-2.0 847.80 11.4 0.00 nan
silex-2.0 397.36 5.3 0.00 nan
symfony-4.0 364.31 4.9 0.00 nan
laravel-5.7 74.44 1.0 0.00 nan
o2system-basic 98.34 1.3 0.00 nan
o2system-hydro 102.01 1.4 0.00 nan
zf-3.0 186.88 2.5 1.15 inf

How to Benchmark

If you want to benchmark PHP extension frameworks like Phalcon, you need to install the extenstions.

Install source code as http://localhost/php-framework-benchmark/:

$ git clone https://github.com/o2system/benchmarks.git
$ cd php-framework-benchmark
$ bash setup.sh

Run benchmarks:

$ bash benchmark.sh

See http://localhost/php-framework-benchmark/.

If you want to benchmark some frameworks:

$ bash setup.sh fatfree-3.5/ slim-3.0/ lumen-5.1/ silex-1.3/
$ bash benchmark.sh fatfree-3.5/ slim-3.0/ lumen-5.1/ silex-1.3/

Linux Kernel Configuration

I added below in /etc/sysctl.conf

# Added
net.netfilter.nf_conntrack_max = 100000
net.nf_conntrack_max = 100000
net.ipv4.tcp_max_tw_buckets = 180000
net.ipv4.tcp_tw_recycle = 1
net.ipv4.tcp_tw_reuse = 1
net.ipv4.tcp_fin_timeout = 10

and run sudo sysctl -p.

If you want to see current configuration, run sudo sysctl -a.

Apache Virtual Host Configuration

<VirtualHost *:80>
  DocumentRoot /home/vagrant/public
</VirtualHost>

References