Multi-dimensional time series-based application server aging model

To observe and study the performance degradation of the application server, client-request programs and server monitoring programs are designed for different scenarios to record different parameters - five categories and 36 parameters altogether. In this paper, primary component analysis method is adopted to reduce dimension, and then multi-dimensional time series analysis method used to set up a model-based on key performance parameter of application server middleware. The statistical result of analyzing the measured data shows that the predicting values derived from the multi-dimensional time series-based application server aging model can match the initial data very well, and that the predicting precision is obviously improved in contrast with the one-dimensional auto regression model. So the aging model may well be adopted for real time predicting of run-time system and its predicting result can be further used as the trigger of system maintenance follow-up action.