An Adaptive Reconfiguration Mechanism for Periodic Software Rejuvenation based on Transient Reliability Analysis

Abstract

While software rejuvenation is used to prevent severe software failures, existing researches generally choose the constant-value periodic policy through steady-state reliability optimization. Since the software reliability declines with the execution time, a steady policy either introduces extra overhead or could not guarantee the reliability constraints. So an adaptive mechanism is proposed to reconfigure the software rejuvenation in the runtime. The transient reliability analysis is used to choose an optimal rejuvenation policy which maintains the software reliability for a certain period of time. A dynamic time series is generated for the reconfiguration process and the optimal rejuvenation policy is re-calculated according to the reconfiguration intervals during the software execution. Experimental studies results show that as the execution time increases, the software reliability drops continuously and the optimal rejuvenation interval should be decreased to maintain the same reliability constraints. This mechanism guarantees the software reliability constraint by resetting the optimal rejuvenation policy dynamically according to a reconfiguration interval time series.