Real-Time Operating Systems

A real-time operating system, or RTOS, manages processor time for software whose correctness includes when a result appears. A control calculation that produces the right value after its deadline is still a failure. The defining property is therefore predictable timing, not high average speed.

An RTOS kernel sits between application tasks and the processor. Tasks contain application work. The scheduler chooses which ready task runs. Interrupt service routines capture urgent hardware events. Kernel objects carry data or coordinate access. Timers turn elapsed time into task activity.

sensor or device interrupt
          ↓
 short interrupt service routine
          ↓ signal, queue item, or notification
      ready task
          ↓ scheduler selects by policy and priority
 bounded application work → output before deadline

The operating system supplies mechanisms. The application still owns the timing design: deadlines, priorities, execution budgets, overload behavior, and recovery actions.

Time is part of correctness

A deadline is the latest acceptable completion time. Latency is the delay from an event to its response. Jitter is variation in that delay across repetitions. Throughput measures completed work per unit time. High throughput does not prove that any particular deadline is met.

Real-time requirements usually fall into three useful classes:

Class	Meaning of a missed deadline	Representative use
Hard real time	The miss violates a system requirement	Protection trip, flight-control update
Firm real time	A late result has no value and may be discarded	Sensor frame used by a current control cycle
Soft real time	A miss reduces quality but service continues	Audio playback or user-interface refresh

These labels describe application consequences. An RTOS does not turn an undocumented timing hope into a guarantee.

Real-Time Operating Systems

Time is part of correctness

Relevant careers