Structural failure models for fault-tolerant distributed computing

Viac o knihe

Errors inevitably occur during system operation despite extensive testing and formal verification. To prevent subsequent failures, critical distributed systems must incorporate fault tolerance, which requires some redundancy, albeit with limitations. Appropriate fault models are essential to define the types and number of tolerable faults in specific contexts. Previous research has often introduced fault models that overlook significant system properties, such as dependent and propagating component failures. Timo Warns presents new structural failure models that are both accurate—covering relevant properties—and tractable for analysis. These models address dependent failures, like those correlated by geographic proximity, and propagating failures, such as those arising from service utilization. To assess the effectiveness of these new models, Warns demonstrates how they can solve key problems in distributed systems with improved resilience and efficiency compared to existing solutions. The comprehensive introduction to distributed systems and the thorough presentation of the new failure models and their evaluations make this work a valuable resource for both researchers and students, serving as an exemplary reference for future software engineering research projects.