Stress Injection Study on Hard Real-Time Operating Systems

The automotive software complexity has increased exponentially in the last 30 years. Nowadays, automotive applications are built on top of hard real-time operating system where many tasks are executed. Due to the automotive high integration levels and the time-to-market, software integration and robustness tests should be performed effectively and efficiently. Infineon Technologies for the AURIX 2G microcontroller has integrated a novel hardware architecture to support the Resource Usage Test and the Stress Test. Despite this hardware support, it has never been used before. Then, it is critical to propose a method to efficiently use this structure and to allow the evaluation of the performance and reliability of the chips. This thesis develops a method and a tool that uses stress injection to analyze the performance, robustness values and boundaries of hard real-time systems under different scenarios. The designer is able: i) to configure the embedded debugging hardware architecture to efficiently explore different stress scenarios; ii) to gather information; and to quantify different types of performance and robustness metrics. The method is automated and fully parameterizable. The developed tool in this thesis is called Galenus, it is integrated into the already existing internal debugging environment of Infineon Technologies for the AURIX microcontroller. The stress injection is based on the reduction of the effective performance of a SoC component (e.g., TriCore within AURIX). The stress injection allows to assess the sensitivity of the SoC under different stress scenarios. These scenarios are defined on the offline initial state using formal methods of scheduling theory. Using the stress injection method, the SoC designer can identify possible risk scenarios testing the performance and robustness of the system at runtime. This thesis is based on the stress injection by CPU suspension within two types of software application, RTOS and Bare-metal.