Overcoming Early-Life Failure and Aging Challenges for Robust System Design

Abstract

The biggest challenge in designing robust systems is to minimize the costs of error detection. Most existing error detection techniques suffer from high power and performance costs, and / or additional design complexity. Circuit failure prediction, together with CASP on-line diagnostics, enable design of robust systems that can effectively overcome reliability challenges associated with early-life failures and aging. The key attractive feature of such an approach is its significantly reduced power cost compared to traditional error detection. It also opens up new research opportunities across multiple abstraction layers (circuit, architecture, virtualization/OS, and applications) for designing optimized robust systems with respect to reliability requirements while balancing power, performance, area, and design complexity constraints. Such global optimization is essential for robust systems of the future.