Rate-Monotonic Scheduling (RMS) in Real-Time Systems: Rule, Bound and Example

Rate-monotonic scheduling (RMS) is a fixed-priority scheduling policy for real-time systems in which tasks with shorter periods are given higher priority. It is the standard way to assign priorities to periodic tasks on a real-time operating system, and it comes with a simple test that tells you, before the system runs, whether every task will meet its deadline.

The problem RMS solves

An embedded controller runs several periodic jobs: read a sensor every 2 ms, run a control loop every 10 ms, update a display every 50 ms, send telemetry every 500 ms. Each must finish before its next instance starts. On a pre-emptive RTOS you assign each task a priority; the question is which priority order guarantees that none of them misses a deadline.

The rule

Assign priorities in order of period: the task with the shortest period gets the highest priority. The 2 ms task runs at the top, then 10 ms, then 50 ms, then 500 ms. For tasks whose deadline equals their period, this ordering is optimal among fixed-priority schemes: if any fixed-priority assignment can meet all deadlines, the rate-monotonic one can.

The schedulability test

For n periodic, independent tasks with computation time C_i and period T_i, define utilisation U = Σ C_i/T_i. The Liu and Layland bound says all deadlines are met if:

U ≤ n (2^1/n − 1)

The bound is sufficient, not necessary: a task set above the bound may still be schedulable, and an exact response-time analysis can confirm it. Below the bound you are guaranteed safe.

Worked example

U = 0.640, below the four-task bound of 0.757, so every deadline is guaranteed under RMS. If the control loop grew to 5 ms (U = 0.89), the bound would be exceeded and a response-time analysis would be needed; in that case the telemetry task would be the one at risk.

Assumptions and what breaks them

• Tasks are periodic and independent. Shared resources protected by mutexes introduce blocking; use priority inheritance or priority ceiling protocols and add the worst-case blocking time to the analysis.
• Deadline equals period. If deadlines are shorter than periods, deadline-monotonic scheduling (priority by deadline) is the right variant.
• Zero context-switch cost. Real switches take microseconds; include them in C_i.
• Known worst-case execution times. Measure or analyse them; optimistic estimates make the whole analysis meaningless.
• Interrupts run above all tasks; treat their worst-case load as the highest-priority “task”.

RMS vs earliest-deadline-first (EDF)

EDF is a dynamic-priority policy that always runs the task with the nearest deadline and can use 100% of the processor. It is optimal in theory but has unpredictable behaviour under overload and is less commonly implemented in commercial RTOSes. RMS is simpler, predictable, degrades gracefully (low-priority tasks miss first) and is what most RTOS schedulers support natively, which is why it dominates in practice.

Applying RMS on a real RTOS

1. List every periodic task with its period and measured worst-case execution time.
2. Sort by period; assign priorities in that order (remember most RTOSes use higher number = higher priority, some the reverse).
3. Compute U and compare with the bound; run response-time analysis if above it.
4. Add blocking times from shared resources and interrupt load.
5. Verify on hardware with timing instrumentation: measure actual response times against the analysis.

Task design, priorities and timing analysis are covered hands-on in our embedded systems course; see embedded operating systems for RTOS fundamentals and characteristics of embedded systems for why real-time constraints matter.

Frequently asked questions

What is rate-monotonic scheduling?

A fixed-priority real-time scheduling policy that gives higher priority to tasks with shorter periods.

What is the RMS utilisation bound?

n(2^1/n − 1), which approaches 69.3% as the number of tasks grows. A task set with total utilisation below the bound is guaranteed schedulable.

Is RMS optimal?

Among fixed-priority policies for periodic tasks with deadlines equal to periods, yes: if any fixed-priority assignment works, RMS works.

What is the difference between RMS and EDF?

RMS uses fixed priorities by period and is simple and predictable; EDF uses dynamic priorities by deadline and can reach full utilisation but behaves less predictably under overload.

What is priority inversion and how does it affect RMS?

A high-priority task blocked on a resource held by a low-priority task that is itself pre-empted by a medium-priority task. It breaks the RMS analysis unless priority inheritance or ceiling protocols are used and blocking time is accounted for.