Sever performance degradation analysis based on average load chaotic time series forecast

A long-running Web software system may lead to the exhaustion of resources, which cause performance degradation. To solve that problem, needs to predict the crucial resources using situation, and then carry out the proper software rejuvenation strategies. At first, this paper identify the average load chaotic character which can be described by using G-P algorithm to analyze correlation dimension changing with embedding dimension, then get the largest Lyapunov exponent through small data method and build chaotic time series prediction model based on largest Lyapunov exponent for average load time series. The experimental results show that the prediction model can precisely make short-time prediction to the Web server´s load, which can efficiently estimate the performance degradation situation and provide foundation for the software rejuvenation.