Reliable Software for Unreliable Hardware

This book demonstrates a highly reliable software system for embedded computing, describing novel concepts, strategies, and implementations to leverage multiple system layers in an integrated fashion for reliability optimization under user-defined, tolerable performance overhead constraints. The authors' approach bridges the gap between hardware and software by quantifying the effects of hardware-level faults at the software level, while accounting for the processor architecture and layout. Readers will learn to leverage multiple software layers to achieve high soft error resilience on unreliable or partially-reliable hardware, while exploiting the inherent error masking characteristics and soft error mitigation potential at different software layers. The authors focus on designing reliable applications, which will execute on an unreliable (or partially-reliable) embedded hardware, given reliability threats such as soft errors, aging, and process variations. Readers will benefit from the techniques described for cross-layer software program reliability modeling and optimization at various granularities (e.g., instruction and function) and at different system design abstractions.§