qbix/webserver

Qbix Server is a web server written entirely in PHP that replaces the usual combination of nginx and php-fpm with a single process. That one process can serve plain static files, run your PHP scripts, handle WebSocket connections, and even show a live dashboard, all without installing or configuring a separate web server. The main idea behind it is memory efficiency. A typical php-fpm setup starts a separate worker process for every concurrent request, and each of those workers loads a full copy of your framework and its configuration into memory, often thirty to sixty megabytes each. Qbix Server instead loads your application's classes and configuration once, then creates new worker processes from that already loaded state using a Unix feature called copy on write, so the shared code only takes up memory a single time. Each new worker then only needs a few megabytes for its own request specific data, which the README claims can let the same server handle around ten times as many concurrent PHP requests on the same hardware. Building an application on top of Qbix Server means writing plain PHP files organized into folders, without needing extra framework code to wire things together. A file under an api folder responding to GET requests becomes an HTTP endpoint, a file under a chat folder can handle WebSocket messages for one connected user with state that persists between messages, and a special room type of file can hold shared state for everyone connected to the same group, like players in an online game. Each of these process types is automatically cleaned up when the connection ends or the room empties, so there is no manual cleanup code required. Getting started is a single command: clone the repository, point it at a folder containing a web page, and run the server with a chosen port. The project also ships a benchmarked comparison against nginx, showing it can be somewhat slower for simple one off requests but noticeably faster once browsers use their normal keep alive connections, and specifically much faster for real PHP workloads because it skips repeatedly reloading the whole framework on every request. The full README is longer than what was shown.