Abstract: Traditional performance monitoring system monitors system resources to reflect and evaluate the system running state indirectly. But this method has greater difference to the system performance perceived by user. For end-users,the direct experience to the system performance is the response time of the request, which is affected by many factors.Based on gene expression programming theory, a gene expression programming (GEP) algorithm was proposed in the paper. Using GEP algorithm, a mathematical model (GEP model) between response time and a variety of system resources was established to evaluate the computing system performance and predict the performance changes. Finally,for specific simulation environment and sampling data, the algorithm was used to obtain the multi-nonlinear model about response time and the results show that the model can better predict the system performance.

Key words: Performance evaluation, Performance monitoring, Gene expression programming, Software rejuvenation