In order to understand the functionality of a real-time operating system (RTOS), you need to know about both general and Linux specific mechanisms.
A real-time system is a platform that is running an RTOS and at least one real-time task. An RT task is a task that must be completed before a certain deadline. These tasks can be periodic or aperiodic have a priority to make sure that they are scheduled according to their frequency of execution and the strictness of their deadline. There can also be non real-time tasks running on the system as well.
The mainline Linux kernel implements three different preemption models for different fields of application like servers or desktop PCs. With the PREEMPT_RT patch two additional preemption models are available. The “Fully Preemptible Kernel” model is the one that turns Linux into an RTOS.
When a task with high priority is blocked by one with low priority because of a mutually exclusive resource, a third task with priority in between the others can run and finish before the task with the highest priority resumes. This phenomenon is known as priority inversion. It can be solved by priority inheritance.
The Linux kernel implements several real-time and non real-time scheduling policies. Depending on the scheduling policy of the tasks the scheduler decides which task is swapped out and which task is processed next.
The RT throttling mechanism prevents a system to hang, if there is a programming failure in a real-time application. This mechanism makes it possible to stop such an application. The settings for RT throttling are exported into the proc filesystem.
Low latency is a critical requirement in real-time computing environments because it ensures that tasks or processes can respond quickly and predictably to external events or inputs.
Dynticks or Dynamic ticks or no HZ mode reduces timer interrupts overhead.