Diskless nodes process data, thus using their own CPU and RAM to run software, but do not store data persistently—that task is handed off to a server. This is distinct from thin clients, in which all significant processing happens remotely, on the server—the only software that runs on a thin client is the "thin" (i.e. relatively small and simple) client software, which handles simple input/output tasks to communicate with the user, such as drawing a dialog box on the display or waiting for user input.