RoadRunner VS Workerman

FrankenPHP

why link to the old URL when linking to its redirected one is less confusing?

https://github.com/roadrunner-server/roadrunner#readme (MIT)

Don't follow any advice to use Apache as a reverse proxy, or bundle php with a classic web server.

There are real application servers using an event loop by now, most notably Roadrunner (https://roadrunner.dev), FrankenPHP (https://frankenphp.dev), Laravel Octane (https://laravel.com/docs/10.x/octane#introduction), Swoole Bridge for Symfony (https://github.com/insidestyles/swoole-bridge-bundle).

In general, you can do a lot with OpenSwoole or Roadrunner. They are vastly superior (in a container scenario) to any other suggestion in this thread!

https://roadrunner.dev which golang php application server which can replace nginx +php fpm has a module for running golang via php using sockets.

You could also use something like EventMachine (In ruby), Twisted (Python), Node (JS) or ReactPHP (for PHP) that will use the language and turn it into a web application server, and then you'll have only one long running process that handle all your requests with shared memory. You could even use something more fancy like RoadRunner in the case of PHP.

Have you considered optimizing the Laravel app? I love Go, but full rewrite for such low rate sounds overkill. Have you looked at the RoadRunner PHP application sample, for example?

Has state resetter for long-running workers serving multiple requests such as RoadRunner or Swoole.

I remember a consulting gig where they were pushing 50k requests per second with PHP in a 2018 laptop by using this framework https://github.com/walkor/workerman

In a real application, with database connection pooling and auth sessions, it went down to 15k requests/s.

And that was PHP7. PHP8 introduced JIT so it's probably significantly faster these days and hopefully fully typed.

The code above is an example of creating a websocket server from the usage example on the workerman/workerman page on github.

Workerman

That's cute. I managed to get 100k requests per second on a single i7 CPU this week using standard blocking code with Workerman. Zero added cognitive load. It also does DB pooling, workers, sockets and consumes relatively lower memory than other solutions.