Development of Real-Time Systems

Here is where it all starts! We will make a brave attempt to start your future career in real-time systems! This week starts by learning the basic building stones in real-time systems and the system parameters required to successfully construct a real-time system. We introduce you to the corner stone of real-time systems, namely the scheduler – and its task in real-time schedules. You learn also what kind of real-time guarantees are needed in which systems. Concretely, you will learn (1) What is needed to create a real-time system (2) Where real-time requirements are needed. (3) The task and job structure and the parameters needed to schedule a task. (4) Difference between pre-emptive and non-pre-emptive tasks. This course is also part of a Blended Master Programme in Embedded Systems.

In this week we start to actually do some scheduling. We firstly have a look at the simplest type of scheduler – the clock driven scheduler. After this we deepen our learning with the fixed priority scheduler and the monotonic schedulers. We learn how the priorities are determined using these schedulers and we learn when/when not to use these schedulers. Secondly we learn how to determine if a system is overloaded in order to validate schedules without complete system simulation.Concretely, after attending this week you will be able to: (1) Schedule a set of tasks with the clock driven scheduler, with the fixed priority scheduler, with the monotonic schedulers. (2) Determine if a system is overloaded using the total-utilization method and the Urm method. (3) Program FreeRTOS to schedule a set of tasks using the fixed priority scheduler

In the previous week we learned the limitations of the total-utilization method and the Urm method. We start this week off by extending these tests to tasks with short response times. The new method is also, as usual, used in practice! We then focus on improving the optimality of real-time schedules. This is done by learning the principles behind dynamic scheduling methods. Two new schedulers using the dynamic scheduling principle is learned and used in practice with a few examples; the LST scheduler and the EDF scheduler.Concretely, you will learn: (1) How to determine feasibility of a set of tasks using the time demand analysis method. (2) How to schedule a set of tasks using LST and EDF. (3) To determine when a dynamic scheduler is appropriate and when it is not appropriate.

All things are easy and nice when not caring about non-periodic jobs. When including non-periodic jobs, the schedulers must take these into account to make a feasible schedule, which we learn this week. The problem is that a non-periodic job can arrive at any time, even if a periodic job is already scheduled. We learn both how to optimize a schedule for non-periodic jobs and how to validate a schedule when non-periodic jobs arrive to the schedule.

Concretely, we will learn: (1) How to use the slack stealing method to optimize a schedule with non-periodic jobs (2) How to use the LRT scheduler to optimize a schedule with non-periodic jobs (3) How to use the deferrable server to optimize a schedule with non-periodic jobs (4) Formally verify a schedule with non-periodic jobs