Software/Hardware RTOS and System Overhead in a Hard Real-Time Environment

Using a Real-Time Operating Systems (RTOS) in an embedded system is a good solution for reducing design time, but an RTOS can affect runtime behaviours. Developers of hard real-time systems should be very careful, as these RTOS effects may result in missing a critical deadline. An RTOS lacks determinism because of associated jitter and system overhead. Therefore, it is important to have a tool that can be used to decide whether or not a system using an RTOS can meet hard real-time constraints.;This thesis extends the standard task response time model to account for system overhead and to improve the prediction of task response times. Furthermore, runtime determinism can be improved by partitioning the RTOS into hardware and software components. A Field Programmable Gate Array (FPGA) is used to build a prototype to explore a partitioning of the muC/OS-II RTOS. A case study illustrates the improved determinism of the partitioned RTOS system, and provides performance data needed to compare the standard response time model to the improved model. The results show that applying the improved model to the partitioned RTOS can improve the response time prediction accuracy by several orders of magnitude.