EFFECTIVE ALGORITHMS TO ESTIMATE THE OPTIMAL SOFTWARE REJUVENATION SCHEDULE UNDER CENSORING
Koichiro Rinsaka, Tadashi Dohi · WORLD SCIENTIFIC eBooks · 2009
AbstractIn this chapter, we consider the optimal software rejuvenation schedule which maximizes the steady-state system availability. We develop statistical algorithms to improve the estimation accuracy in the situation where randomly censored failure time data are obtained. More precisely, based on the kernel density estimation, we estimate the underlying failure time distribution. We propose the framework based on the kernel density estimation to estimate optimal software rejuvenation schedules from censored sample data. In simulation experiments, we show the improvement in the convergence speed to the real optimal solution in comparison with the conventional algorithm.