Understanding fault-tolerance vulnerabilities in advanced SoC FPGAs for critical applications

Abstract

The emergence of heterogeneous FPGA-based SoCs and their growing complexity fueled by the introduction of various accelerators bring the reliability aspect of these systems to the front. The concern is particularly important for critical applications, such as safety-critical autonomous vehicles, real-time systems, space missions, health, and security cyber–physical systems. This paper presents an analysis of the most common types of errors caused by faults in different components and identifies the most critical and vulnerable components based on the literature survey. The study looks into vendor-provided and user-space reliability solutions. The paper highlights the existing fault-tolerance gaps in the modern SoC FPGAs and outlines a vision for future solutions.