Why PHP-FPM performance is better than Octane?

[liorkgit]

I setup a clean Laravel 9 project.
I then setup Octane with RoadRunner.

I run it on a VirtualBox VM (Ubuntu 22.04) in a Windows 11 host.

My PC:
CPU: Ryzen 5 3600
RAM: 32GB - 2x16GB DDR4 3200Mhz CL16
Storage: Samsung 970 Evo (Not plus), 500GB

The VM:
CPU: 4 Cores
RAM: 4GB
Storage: 10GB fixed

I tested and compared the performance between PHP-FPM and Octane with nginx ,using wrk:
https://github.com/wg/wrk

Ran the benchmark on Laravel's default homepage

These are the nginx config files for each of the setups:

1. The Octane setup:

map $http_upgrade $connection_upgrade {
    default upgrade;
    ''      close;
}

server {
    listen 80;
    listen [::]:80;
    server_name myapp.dev;
    server_tokens off;
    root /var/www/html/myapp/public;

    index index.php;

    charset utf-8;

    location /index.php {
        try_files /not_exists @octane;
    }

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

    location = /favicon.ico { access_log off; log_not_found off; }
    location = /robots.txt  { access_log off; log_not_found off; }

    access_log off;
    error_log  /var/log/nginx/myapp.dev-error.log error;

    error_page 404 /index.php;

    location @octane {
        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://127.0.0.1:8000$suffix;
    }
}

2. The PHP-FPM setup:

server {
    listen 80;
    server_name myapp.dev;
    root /var/www/html/myapp/public;

    add_header X-Frame-Options "SAMEORIGIN";
    add_header X-XSS-Protection "1; mode=block";
    add_header X-Content-Type-Options "nosniff";

    index index.html index.htm index.php;

    charset utf-8;

    location / {
        try_files $uri $uri/ /index.php?$query_string;
    }

    location = /favicon.ico { access_log off; log_not_found off; }
    location = /robots.txt  { access_log off; log_not_found off; }

    error_page 404 /index.php;

    location ~ \.php$ {
        fastcgi_pass unix:/var/run/php/php-fpm.sock;
        fastcgi_index index.php;
        fastcgi_param SCRIPT_FILENAME $realpath_root$fastcgi_script_name;
        include fastcgi_params;
    }

    location ~ /\.(?!well-known).* {
        deny all;
    }
}

The average req/sec when using PHP-FPM was higher than when using Octane. This is how I ran wrk:

wrk -t4 -c400 -d30s http://127.0.0.1:8080/index.html

I also ran octane with 4 workers:

php artisan octane:start --workers=4

The average with PHP-FPM was ~480 req/sec, and the average with Octane was ~400 req/sec

Example benchmarks:

1. PHP-FPM:

user@server:~$ wrk -t4 -c400 -d30s http://127.0.0.1/
Running 30s test @ http://127.0.0.1/
  4 threads and 400 connections
    Thread Stats       Avg             Stdev         Max        +/- Stdev
    Latency            723.95ms      83.16ms      1.04s          81.75%
    Req/Sec          143.19           99.21          525.00        67.24%
16381 requests in 30.10s, 293.92MB read
Requests/sec:    544.13
Transfer/sec:      9.76MB

2. Octane:

user@server:~$ wrk -t4 -c400 -d30s http://127.0.0.1/
Running 30s test @ http://127.0.0.1/
  4 threads and 400 connections
    Thread Stats       Avg             Stdev         Max        +/- Stdev
    Latency          887.53ms      299.89ms      1.53s          61.66%
    Req/Sec          138.80           134.46          782.00       86.98%
13340 requests in 30.04s, 238.36MB read
Requests/sec:    444.09
Transfer/sec:      7.94MB

3. A test run with Laravel's built-in development server:

user@server:~$ wrk -t4 -c400 -d30s http://127.0.0.1:8080/
Running 30s test @ http://127.0.0.1:8080/
  4 threads and 400 connections
    Thread Stats       Avg             Stdev           Max        +/- Stdev
    Latency           767.54ms      626.07ms      1.78s          79.41%
    Req/Sec             44.60           45.79            257.00       87.16%
  2075 requests in 30.04s, 37.04MB read
Socket errors: connect 0, read 2075, write 0, timeout 1973
Requests/sec:     69.08
Transfer/sec:      1.23M

Why?

[Sophist-UK]

Did you try Octane with FPM?

[liorkgit]

Oh I should? Do you know how I should then edit the above nginx config to do it? I am not the best with it and I used the example in the docs: https://laravel.com/docs/9.x/octane#serving-your-application-via-nginx

[Sophist-UK]

Hmm. A bit of research suggests I am talking rubbish - PHP FPM & nginx Unit might be possible to combine (they use different technologies to have faster initialisation of the Laravel framework on each transaction), but Octane uses a different technique to remain running and be pre-initialised to receive web traffic.

Have you seen this page https://habr.com/en/post/646397/ or this one https://stackoverflow.com/questions/73883958/why-php-fpm-performance-is-better-than-octane ?

Speaking from a position of absolute ignorance of actual experience of any of these specific technology, and using only several decades of more general systems programming and performance optimisation experience...

• If the performance of FPM is the same as nginx Unit, I would tend to prefer FPM as it is less specific to nginx and will have more users.
• If the performance of FPM is the same as Octane, I would tend to prefer FPM as you are unlikely to run into Octane-specific issues resulting from reusing the same Laravel initialisation multiple times.