Everything between your keystroke and the pixel on screen.
You press a key on your keyboard. Characters appear on a screen. Between those two events, thousands of operations occur: electrical signals travel from the keyboard to a USB controller, get translated into a scan code, routed through a driver, interpreted by a program, rendered as pixels by a graphics card, and pushed to a display. You never see any of it.
That invisible machinery is the operating system (OS).
An OS does three things:
Your phone has an operating system too — iOS or Android. What happens when you switch between apps? Think about what the OS is doing behind the scenes when you swipe from one app to another.
The OS has layers. At the center is the kernel — the code that runs with full hardware access. It handles memory allocation, process scheduling, device drivers, and system calls. Everything else (your file manager, your web browser, your terminal) runs in user space, where it must ask the kernel for permission to do anything involving hardware.
When a program wants to read a file, it doesn't talk to the disk directly. It makes a system call — a formal request to the kernel. The kernel checks permissions, locates the data on disk, copies it into the program's memory, and returns control. This happens millions of times per second on a busy server.
Why does this matter for datacenter operations? Because when something goes wrong — a disk fails, a process consumes all available memory, a network interface drops — the symptoms you see and the tools you use all interact with the kernel. Understanding this layer is how you move from "the server is broken" to "the kernel remounted the filesystem read-only because it detected I/O errors on the NVMe drive."
If a program crashes, why doesn't it crash the whole computer? What part of the OS prevents that?