Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This licentiate thesis by Dimitar Nikolov focuses on optimizing fault tolerance in real-time systems (RTSs), which are critical in producing correct results within strict time constraints. As modern RTSs are increasingly prone to soft errors, fault tolerance mechanisms like Roll-back Recovery with Checkpointing (RRC) are necessary to detect and recover from such errors, albeit at the cost of time overhead. The thesis introduces the Level of Confidence (LoC) metric to evaluate deadline adherence in both soft and hard RTSs, allowing designers to optimize system performance. It presents mathematical models to determine the optimal number of checkpoints to minimize Average Execution Time (AET) and maximize LoC, alongside techniques for estimating error probability during runtime. Supported by European and Swedish research initiatives, the work addresses significant challenges in computer science and engineering, targeting researchers and professionals in the field.

Fault-tolerant computing. --- Real-time data processing. --- Fault-tolerant computing --- Real-time data processing