Evaluating the data integrity of memory systems by configurable Markov models

In this paper, a novel method for the evaluation of the bit error rate (BER) as measure for assessing data integrity in memory systems is proposed; such method improves modeling by introducing configurability features in the Markov chains to account for environmental and operational changes. For modeling erasures and random errors, the occurrence of new time-varying features is introduced in the analysis to characterize the behavior of memory systems for space applications (using Reed-Solomon codes as EDAC). Moreover, differently from existing techniques, the nature of these features (such as scrubbing and the effects of the so-called South Atlantic Anomaly on SEU rates) is assessed using a deterministic framework.