quillstack/benchmark

Bash/PHP script to benchmark HTTP requests and command line scripts.

1.0.7 2020-09-29 22:28 UTC

This package is auto-updated.

Last update: 2024-05-09 17:08:03 UTC


README

StyleCI Build Status CodeFactor Downloads Packagist PHP Version Support Packagist License

This repository contains a script to test HTTP GET requests or command line calls.

PHP usage

You can install this package using Composer:

composer require --dev quillstack/benchmark

In PHP console you can use this library by running commands:

List available commands

If you installed this library as a package in your project, you can run these commands:

List of available commands
./vendor/bin/benchmark
HTTP GET requests
./vendor/bin/benchmark benchmark:http:get https://example.org 10 2
Command line calls
./vendor/bin/benchmark benchmark:console "php ../test.php" 10 2

Cloned reposotiry

If you cloned this repository to your local computer, use these commands:

./bin/local
./bin/local benchmark:http:get https://example.org 10 2
./bin/local benchmark:console "php ../test.php" 10 2

To see detailed descriptions for every command, ready Bash usage below.

Bash usage

You can also use Bash commands to run benchmarks. These commands work only if you clone this repository to your computer or server.

HTTP GET requests

Usage:

./http_get.sh https://example.org 10 2

Where:

  • 10 is a total number of requests
  • 2 is a number of concurrent requests

Output:

10 requests, 2 concurrently
URL https://example.org
--------------------------------------------------------------------
Took 2.468000 s, 4.051864 requests per second, 0.469415 avg req time

What means we sent 10 GET requests to this host. We decided to send two requests at the time. The entire test took around 2.5 seconds. It means the website could server 4 requests per second with 0.5 seconds the average response time.

Command line calls

If you want to test command line calls, be sure every call respond with an execution time (in seconds):

0.001699

For PHP the script could look like:

<?php

require __DIR__ . '/../vendor/autoload.php';

$start = microtime(true);

$container = new Container();
$service = $container->get(ExampleService::class);

$time = microtime(true) - $start;
$roundedTime = round($time, 6);

echo $roundedTime . PHP_EOL;

Usage example:

./command_line.sh "php ../test.php" 10 2

Output:

10 calls, 2 concurrently
Command `php ../test.php`
-------------------------------------------------------------------
Took 0.247000 s, 40.485830 calls per second, 0.001087 avg call time

The results say we called the script 10 times with 2 concurrently calls. Our test took around 250 milliseconds, what means we could call this script 40 times per seconds, and average call time would be 1 millisecond.

Different results

You can have different results for different parameters. For example the results for the same script:

./command_line.sh "php ../test.php" 1000 100

are different for these parameters:

1000 calls, 100 concurrently
Command `php ../test.php`
-------------------------------------------------------------------
Took 7.760000 s, 128.865979 calls per second, 0.080927 avg call time

Because we increased the number of concurrent calls.

If we use the same concurrent calls like in the first command line example:

./command_line.sh "php ../test.php" 1000 2

The results should be similar to the first ones:

1000 calls, 2 concurrently
Command `php ../di/public/index.php`
-------------------------------------------------------------------
Took 22.795000 s, 43.869270 calls per second, 0.001178 avg call time

What gives use around 40 calls per second, if we have two concurrent calls.

Quill Stack

If you want to know more about other solutions, visit the website:
https://quillstack.com/

The Quill Stack