Abstract

Software aging results from runtime environment degradation, and is significantly correlated with available computing resources. A set of variables evolving with time can describe the running state of computer system. Consequently, software aging is analogous to evolution of a dynamic system in this paper. We construct a nonlinear dynamic model based on the experimental observations. First, we assume the mathematical form of nonlinear dynamic equations. Then, we select resource parameters which can reflect the “health” of the whole computer system as variables of our model. Finally, we estimate the values of each parameters in our model using nonlinear inversion. Our approach is validated by two different datasets. The dynamic model can describe the evolution of software aging and interpret the interplay of various resource parameters. Moreover, this model can be used to forecast abrupt state degradation and help us to explore the root cause of software aging. For example, by comparing the output of our model against real values, with a suspected “aging factor” as input, we can identify which resource variable is the root cause of injuring the stability of computer system.