With rapid advances in computer memory capacity and performance, coupled with corresponding increases in the expense of field service calls, memory reliability and optimal maintenance strategies have become more and more important in terms of customer satisfaction and field service cost. At the same time, significant improvements in error correction and recovery over recent years have made the prediction of uncorrectable error and field service frequency much more difficult. This paper describes a Monte Carlo simulator which can predict uncorrectable error rates and field-replaceable-unit replacement rates for a wide range of memory architectures and under a variety of maintenance strategies. The model provides valuable information for performing sensitivity studies of intrinsic failure rates for memory components, for performing tradeoff studies of alternative storage module and card organizations, for evaluating system functions, and for establishing optimum maintenance strategies.
PDF