Number of Events and Time to Failure Distributions for Error Correction Protected Memories

Designing the memory system for an application has always been a complex task. This complexity has been increasing in the last years due the technology scaling, since memories are becoming more sensitive to soft errors. This issue is aggravated when the application is running in a harsh environment, e.g., in the presence of intense radiation sources. Therefore, the task to determine the optimal memory size and optimal protection mechanism is a challenge for designers to provide appropriate fault tolerance. In this paper, some classical reliability parameters, as the mean time to failure and the mean number of events to failure are revisited, but extending their scope with their probabilistic distributions. These can be of great help to designers when analyzing the consequences on reliability of certain decisions related to the memory size and protection codes.