Software aging estimation and prediction of a real VOD system based on PCA and neural networks

The phenomenon of software aging refers to the exhaustion of operating system resource, fragmentation and accumulation of errors, which results in progressive performance degradation or transient failures or even crashes of applications. In this paper, we investigate the software aging patterns of a real VOD system. First, we collect data on several system resource usage and application server. Then, non-parametric statistical methods and linear regression models are adopted to detect aging and estimate trends in the data sets. Finally, artificial neural network (ANN) models are constructed to model the extracted data series of systematic parameters and to predict software aging of the VOD system. In order to reduce the complexity of ANN and to improve its efficiency, principal component analysis (PCA) is used to reduce the dimensionality of input variables of ANN. The experimental results show that the software aging prediction model based on ANN is superior to the time series models in the aspects of prediction precision. Based on the models employed here, software rejuvenation policies can be triggered by actual measurements.