Evaluation of software performance affected by aging

A number of studies have reported the phenomenon of “Software aging”, characterized by progressive software performance degradation. This is mainly caused by the exhaustion of the combination of system resources. Traditionally, modeling and analysis of software aging has been restricted to a single resource aging. The combined effect of software aging due to the interaction of several resources remains to be explored. A BP network model for the evaluation of software performance affected by aging is proposed in this paper, and then a robust locally weighted regression algorithm is presented to identify the inflexion point. The shape of performance curve changes greatly at the inflexion point, which makes it easy to determine when to perform software rejuvenation. Experimental results based on actual data sets verify the proposed model effectiveness in evaluation of the software performance.