Data Age Formulation and Analysis in Real-Time Embedded Systems - Fault tolerant and Thermal aware perspectives

Abstract

Safety-critical real-time systems demand meticulous attention to both temporal requirements and fault tolerance mechanisms. This thesis addresses critical challenges in analyzing and optimizing system performance under the influence of transient faults and thermal constraints. Our first work introduces an analytical framework for evaluating data age in real-time task chains augmented with checkpointing mechanisms. Transient faults pose significant threats to system predictability, and conventional fault tolerance methods such as checkpointing can exacerbate data age dynamics. Our framework offers a systematic approach to quantifying data age in both single-core and multi-core platforms, validated through extensive simulations. The results demonstrate that the data age bounds achieved are comparable to those of existing techniques, while the computational overhead remains minimal in relation to traditional methods. This balance enhances the predictability and reliability of safety-critical systems, ensuring efficient performance without sacrificing accuracy or increasing resource consumption significantly. The second work focuses on the analysis to incorporate thermal constraints alongside data age considerations in task scheduling for safety-critical applications. By introducing a thermalaware data age analysis framework, this work addresses the pressing need for simultaneous optimization of task periods to ensure both data age and processor thermal safety. Through systematic evaluations, we establish the efficacy of our approach in deriving schedulable task periods, thus mitigating risks associated with invalid data consumption and thermal violations in safety-critical systems. Results underscore the effectiveness of our framework in achieving optimal task scheduling while meeting stringent safety and thermal requirements. Together, these contributions provide essential insights and methodologies to enhance the resilience and performance predictability of safety-critical real-time systems