Wavelet and Neural Network Based Approach for Software Aging Prediction
Liu Feng-yu · Jisuanji fangzhen · 2006
A number of recent studies have reported the phenomenon of “software aging”, characterized by progressive performance degradation and a sudden crash/hang of a software system due to exhaustion of operating system resources, fragmentation and accumulation of errors. To counteract this phenomenon, a novel four-stage method for software aging forecast in operation was proposed. The prior data of software performance parameters were treated as time series. The forecast method combines wavelet multiresolution decomposition and neural networks. First, a smoothing unit was applied based on the wavelet multiresolution analysis to reduce the influence of noise. Second, the special performance data was decomposed into different scales by non-decimated Haar wavelet decomposition. Third, each scale was predicted by a separate neural network. The weights and biases were initialized by an algorithm based on immunology inspiration and simulated annealing, which improved the prediction results as compared to random initialization. Last, the next sample of the original time series was predicted by another neural network. The proposed method was tested using the performance parameters data collected from a realistic software system to evaluate the forecasting performance.