Preemptive Scheduling: The Foundation of RTOS Responsiveness

Principle: The scheduler has the authority to forcefully halt (preempt) a currently running task if a task of higher priority transitions to the Ready state (e.g., an interrupt signals data arrival, unblocking a high-priority processing task). This immediate preemption ensures that the most critical tasks gain CPU access without delay.

Advantages:
- Guaranteed Responsiveness: High-priority tasks respond to events with minimal and predictable latency.
- Optimality for Urgency: Best suited for systems where certain tasks are genuinely more time-critical than others.

Disadvantages:
- Increased Complexity: Requires careful handling of shared resources using synchronization primitives to prevent data corruption.
- Context Switching Overhead: Incurs overhead for saving and restoring task contexts.

Priority-Based Preemption: The most prevalent form in RTOS. Each task is assigned a numerical priority (e.g., 0-255, where lower numbers might indicate higher priority or vice-versa, depending on the RTOS convention). The scheduler continuously ensures that the task currently in the Running state is always the highest-priority task among all those currently in the Ready state. If a new task becomes ready with a higher priority than the currently running task, a context switch occurs immediately.

Non-Preemptive (Cooperative) Scheduling: Simpler, Less Predictable

Principle: A task, once it begins executing, will continue to run without interruption until it voluntarily relinquishes control of the CPU. This happens only when the task:
- Explicitly calls a yield function (e.g., task_yield()).
- Enters the Blocked state (e.g., waiting for a delay, a message, or a resource).

Advantages:
- Simpler Implementation: Reduces the complexity of the scheduler and eliminates the need for some of the more complex synchronization primitives (though race conditions can still occur if not careful).
- Lower Context Switching Overhead: Context switches occur less frequently and only at predictable points initiated by the tasks themselves.

Disadvantages:
- Poor Responsiveness for High-Priority Tasks: A low-priority task that fails to yield or blocks can indefinitely monopolize the CPU, causing high-priority tasks to miss their deadlines.
- Not Suitable for Hard Real-Time Systems: Lacks the necessary guarantees for critical applications where timing is paramount.

Crucial Real-Time Scheduling Algorithms (for Preemptive Systems)

These algorithms are fundamental to understanding how an RTOS ensures deadlines are met.

Rate Monotonic Scheduling (RMS):
- Category: A static-priority (priorities are assigned once at design time and do not change), preemptive scheduling algorithm specifically designed for periodic tasks.
- Priority Assignment Rule: The core rule of RMS is simple: tasks with shorter periods (i.e., tasks that need to run more frequently) are assigned higher priorities. Conversely, tasks with longer periods get lower priorities. The intuition is that tasks with tighter deadlines (more frequent execution) are more urgent.
- Optimality (for Fixed-Priority): RMS holds a significant theoretical property: it is considered optimal among all fixed-priority scheduling algorithms for a set of independent, periodic tasks on a single processor. This means if a set of such tasks can be scheduled by any fixed-priority algorithm without missing deadlines, then it can also be scheduled by RMS.
- Schedulability Test (Liu and Layland Utilization Bound): A key analytical tool for RMS. For a set of n independent periodic tasks, a sufficient (though not necessary) condition to guarantee schedulability (i.e., all tasks will meet their deadlines) is that the total CPU utilization (U) must be less than or equal to a specific bound: U=∑i=1n (Ci /Ti )≤n(21/n−1) Where:
- Ci represents the Worst-Case Execution Time (WCET) of task i.
- Ti represents the period of task i.
As the number of tasks (n) approaches infinity, this bound converges to ln(2)≈0.693 (or approximately 69.3% CPU utilization). This means if your tasks utilize more than about 69.3% of the CPU, RMS cannot guarantee schedulability.

Earliest Deadline First (EDF) Scheduling:
- Category: A dynamic-priority (priorities change at runtime), preemptive scheduling algorithm.
- Priority Assignment Rule: At any given moment, the task with the earliest absolute deadline is always assigned the highest priority and executed. As tasks run and new tasks become ready, their deadlines are compared, and priorities adjust accordingly.
- Optimality (for Preemptive): EDF is considered optimal among all preemptive scheduling algorithms for a set of independent tasks.
- Schedulability Test: For any set of tasks, EDF can schedule them without missing deadlines if their total CPU utilization is ≤100%.

Round-Robin Scheduling

Category: A time-sliced, preemptive algorithm (often used within tasks of the same priority level).

Principle: Tasks are given a fixed time quantum (or "time slice"). When a task's quantum expires, it is preempted, and the next task in the ready queue (usually of the same priority) gets the CPU for its quantum. This repeats in a circular fashion.

Usage: Frequently used for tasks that have the same priority level in a multi-priority RTOS system, ensuring fairness among them. Can also be used as a simple non-real-time scheduler for less critical systems.

Deterministic Properties: While it ensures fairness, its real-time guarantees are weaker than RMS or EDF unless the time quantum is carefully chosen relative to task deadlines.