Recently, a kind of phenomenon known as software aging has been reported in many studies, characterized by progressive performance degradation or abrupt crash/hang. Software aging is directly related to resources usage. In this paper, we set up an experimental platform with multi-clients, collect data on system resource usage on web server and log data on performance on client. It is observed that there is performance degradation while the amount of available memory remains constant. Results show that there is significant correlation between response time and buffer. Finally, we calculate the largest Lyapunov exponent of time series of averageload, and results show that software aging process is chaotic.