Managing resources for large-scale testing

This post introduces One World, a new unified resource management system for large-scale testing at Facebook. It details the architecture, including runtime worker services, the One World daemon, and the Jupiter scheduler. It explains how One World manages tens of thousands of resources (mobile devices, browsers, emulators) and executes over a million jobs daily. The post highlights the challenges of device reliability and the solutions implemented, such as state machines for worker services, health checks, and state restoration. It also describes the mechanisms for remote access to Android (via adb tunnels) and iOS (via FBSimulatorControl and WebDriverAgent) devices, and the role of the One World daemon in session management and resource leasing. The satellite mode for connecting local devices to the global system is also presented.