The need for real-time scheduling

A real-time embedded system is a system with a dedicated purpose. The real-time system may operate standalone or it may be a component or subsystem of a larger device. Real-time systems are often event-driven and must produce the same output and timing when given the same initial conditions. A real-time system might be built using a microcontroller system that uses a bare-metal scheduler or a real-time operating system (RTOS) to schedule all of its system tasks. Alternatively, it could be built using a System on Chip (SoC) or Field Programming Gate Array (FPGA).

Every embedded system is not necessarily a real-time system. An application processor such as Raspberry Pi using Raspbian or Linux would not be a real-time system because, for a given set of inputs, while the system may give the same output, the time taken can vary wildly due to the multitasking nature of the system. General-purpose operating systems often interrupt tasks to handle OS-related functions, which results in the computing time being variable and non-deterministic.

There are several characteristics that can be used to identify a real-time embedded system:

• They're event-driven as they do not poll inputs.
• They're deterministic because when given the same initial conditions, they produce the same outputs in the same time frame.
• They're resource-constrained in some manner; for example, clock speed, memory, or energy consumption.
• They use a dedicated microcontroller-based processor.
• They may use an RTOS to manage system tasks.

Real-time system types

Real-time systems can be subdivided into two categories: soft real-time and hard real-time systems. Both types require that the system executes in a deterministic and predictable manner. However, they differ in what happens if a deadline is missed. A soft real-time system that misses a deadline is considered to be annoying to its users. It's undesirable for the deadline to be missed and may decrease the usefulness of the system after the deadline, but it's not critical. A hard real-time system, on the other hand, will dramatically decrease its usefulness after a deadline and results in a fatal system fault.

An example of a soft real-time system is a Human Machine Interface (HMI) with a touch controller that is controlling a home furnace. There may be a deadline where the system needs to respond to user input within 1 second of the screen being touched. If a user goes and touches the screen but the system doesn't respond for 3 or 4 seconds, the result is not world ending, but it may make the user complain about how slow the system is.

A hard real-time system could be an electronic braking system that needs to respond to a user pressing the brake pedal within 30 milliseconds. If a user were to press the brake and it took 2 seconds for the brakes to respond, the outcome could be critical. The system's failure to respond could result in injury to the user and dramatically decreases the usefulness of the embedded system.

It is possible to have an embedded system that has a mix of hard and soft requirements. The software in an embedded system is often subdivided into separate tasks based on function and timing requirements. We might find that the user interface on a system is considered to have soft real-time requirements, while the actuator control task must have hard real-time requirements. The type of system that is being built will often factor in the type of scheduler that is used in the solution.

Now, let's explore the different scheduling architectures that can be used with MicroPython to achieve real-time performance.