The Risk of Data Corruption in Microprocessor
Robert W. Horst, D. Jewett, Daniel E. Lenoski · 1993
Most microprocessor-based systems provide little or no checking for failures of the microprocessor itseIf. These systems rely on the high reliability of VLSI and the inherent run-time checks (e.g., virtual memory protection and illegal instruction exceptions) to ensure data integrity. In this paper we present a model for analyzing the vulnerability of systems to undetected data corruptions. The model estimates expected corruption rates based on commercial reliability studies, jield experience with fault-tolerant systems and academic fault-injection studies. We jind that without lockstep, an undetected data corruption may occur nearly once per month in a population of 10,OOO processors. Given the available data, however, it is dificult to provide tight bounds for the ftequency of data integrity violations and our current estimates of the data corruption rate vary by more than two orders of magnitude. Further study on transient failures and the impact of faults is required to narrow the range of our predictions and reduce the overall risk of undetected data corruptions.