Quick Laravel Octane test - UPDATED

test-octane.md

Intro

I wanted to quickly take Laravel Octane beta for a spin, as we have been using Swoole for some time and wanted to see how official support performs vs what we have.

Test 2: Forge server

Next test is more real world as we now have a web server and database server within a VPC.

Using Laravel Forge

• App server: PHP 8.0.3, AWS t3.small - 2 vCPUs, 2 GB RAM
• Database: AWS RDS MySQL 8.0 db.t3.micro, 2 vCPUs, 1 GB RAM - using private VPC
• SSL certificate

We will use opcache, cache route and config in Laravel, and adjust settings for performance.

We test for 3 min, using 8 threads and 200 connections each.

PHP-FPM Settings

pm = dynamic
pm.max_children = 100
pm.start_servers = 20
pm.min_spare_servers = 20
pm.max_spare_servers = 40

Nginx - Swoole

upstream swoole-http {
    server 127.0.0.1:8000;
}

map $http_upgrade $connection_upgrade {
    default upgrade;
    ''      close;
}
server {
    listen 443 ssl http2;
    listen [::]:443 ssl http2;
    server_name octane.example.com;
    server_tokens off;
    root /home/forge/octane.example.com/public;
    index index.php;
    charset utf-8;

    # FORGE CONFIG (DO NOT REMOVE!)
    include forge-conf/octane.example.com/server/*;

    location = /index.php {
        try_files /not_exists @swoole;
    }

    location / {
        try_files $uri $uri/ @swoole;
    }

    location @swoole {
        set $suffix "";

        if ($uri = /index.php) {
            set $suffix ?$query_string;
        }

        proxy_http_version 1.1;
        proxy_set_header Host $http_host;
        proxy_set_header Scheme $scheme;
        proxy_set_header SERVER_PORT $server_port;
        proxy_set_header REMOTE_ADDR $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection $connection_upgrade;

        proxy_pass http://swoole-http$suffix;
    }
}

• wrk -t8 -c200 -d3m https://testdomain.com/api/ping
• wrk -t8 -c200 -d3m https://testdomain.com/api/users
• wrk -t8 -c200 -d3m https://testdomain.com/api/users/1

Route	Nginx to PHP-FPM on AWS t3.small, RDS db.t3.micro	Nginx to Octane Swoole on AWS t3.small, RDS db.t3.micro
/api/ping	295 req/sec, 98% App CPU	832 req/sec, 84% App CPU
/api/users	77 req/sec, 85% App CPU, 88% DB CPU, 90 DB conn	113 req/sec, 29% App CPU, 54% DB CPU, 4 DB conn
/api/users/:id	189 req/sec, 91% App CPU, 10% DB CPU, 16 DB conn	453 req/sec, 61% App CPU, 6% DB CPU, 2 DB conn

Test 1 - Setup

• Macbook Pro 2019
• Laravel 8.35.1
• Laravel Octane 0.1.0 beta
• Laravel Sail
• Wrk

Basic Laravel app with 5000 users seeded into the database. Only installed Sail and Octane packages. Otherwise 100% vanilla setup.

Test Routes

Route::get('/users', function () {
    return response()->json(User::query()->paginate());
});

Route::get('/users/{userId}', function (int $userId) {
    return response()->json(User::query()->findOrFail($userId));
});

Route::get('/ping', function() {
    return response()->json(['message' => 'pong']);
});

Route	Response	MySQL
/api/ping	Returns a string	No
/api/users	Returns a paginated list of users	Yes
/api/users/:id	Returns a user	Yes

• wrk -t4 -c50 -d30s http://localhost/api/ping
• wrk -t4 -c50 -d30s http://localhost/api/users
• wrk -t4 -c50 -d30s http://localhost/api/users/1

Test Results

All the tests have been conducted on my MacBook Pro 2019 using Docker. Im sure using native would yield better performance. However, we care about the difference - not so much about the actaul numbers.

Tests

1. Nginx to PHP-FPM: classic nginx to fpm setup using docker
2. Nginx proxy to Swoole. Using Laravel-swoole package and nginx
3. Laravel Octane beta and Sail

Route	Nginx to PHP-FPM	Nginx to Swoole	Octane Swoole HTTP
/api/ping	131 req/sec	369 req/sec	766 req/sec
/api/users	82 req/sec	152 req/sec	201 req/sec
/api/users/:id	104 req/sec	290 req/sec	545 req/sec

All the tests have been conducted on my MacBook using Docker. Im sure using native would yield much better performance. However, we care about the difference - not so much about the actaul numbers.

Again, very quick tests - but for now it looks promising.

And about memory usage ?

Fantastic comparison!! A number of interesting observations with your reported results.

• I was surprised by the DB CPU and DB connection count between the two environments. Curious how this work look with some larger DB resources.
• I am curious to know the throughput stats of your Octane setup when you max out CPUs for your APP and DB. Aka - Apples to Apples(ish) resource comparison.