Performance degradation model of application server system based on time series analysis

Abstract

Many kinds of client request applications and server programs are designed for the aging of JUFrame application sever, in which 36 parameters in 5 groups are recorded. The key index 'memory usage' that affects the performance of middleware of application server is modeled based on time series analysis. The results that the auto correlation function (ACF) is clearly tail-dragged and the partial auto correlation function is clearly tail-truncated can be drawn via detailed statistical analysis of sampled data. The result of model simulation shows that the predicted value fits the original data well and can be regarded as the basis of system rejuvenation, and is applicable for predicting runtime systems. Compared to other software aging prediction methods, the proposed method is more suitable for the practical case of systems, and the availability of the system is improved.