Optimal control of latent fault accumulation

Abstract

The problem of controlling latent fault accumulation in digital memory devices in a dynamic space environment is analyzed. Previous analytical modeling work has examined constant error rate environments and constant rate error scrubbing and has generally assumed single error correction. This work is extended to consider environments that cause orders of magnitude variation in the device error rate over the mission life. Such variation can be caused by solar flare activity or time-dependent variations in geomagnetic shielding. Based on the probability models and approximations developed, optimal time-invariant and time-varying error scrubbing policies are analyzed. The models are generalized to allow multiple error correction. Definitions of optimal scrubbing policies are given, and this class of policies is shown to bound the performance of nonoptimal policies.<>