Dynamic Priority Assignment in FreeRTOS Kernel for Improving Performance Metrics

Real-time operating systems typically have support for static task scheduling algorithms only. However, real-time applications often require dynamic scheduling algorithms in order to meet their timing constraints. Furthermore, in the case of overloaded conditions, i.e., scenarios when there is no feasible schedule, the total tardiness of real-time tasks can be significantly decreased by using a custom dynamic priority assignment. This paper presents a method for integrating custom scheduling heuristics to an open source real-time operating system (FreeRTOS). The FreeRTOS task management subsystem was modified in order to support dynamic task priority assignment. The method was evaluated by simulating the execution of periodic preemptable task sets and the results were compared with the default FreeRTOS scheduler with respect to the total weighted tardiness and weighted number of late jobs measure. The system performance in both underloaded and overloaded conditions was improved by using custom dynamic scheduling policies.

I. Introduction

Reliability of a real-time system depends not only on the functional correctness of its components, but also on the timeliness and predictability of task execution [1]. There are many mechanisms implemented in a real-time kernel which influence the predictability of the system. However, the scheduling algorithm is certainly the most important factor, since it is responsible for satisfying timing and resource contention requirements [2].