摘要： 应用小波分析和统计分析相结合的方法来处理系统资源使用相关的数据，实现了在线的软件衰退检测和资源消耗趋势估计。设计一种基于Haar 小波分析的数据平滑技术消除噪声，使用统计分析技术来分析预处理后的数据，以检测衰退的出现，估计资源消耗的趋势斜率，给出资源到达预警值的时间，确定系统的关键性资源。使用真实环境采集的数据进行实验，实验结果表明该方法能有效地检测出软件衰退并给出每一种资源耗尽的估计时间，为恢复策略的制定提供了依据。

关键词: 软件衰退；小波分析；统计分析

Abstract: This paper proposes a hybrid method based on wavelet analysis and known statistical techniques to analyse the usage of system resources, to detect the existence of software aging and estimate the time to exhaustion. It proposes a methodology based on Haar wavelet analysis to denoise. Then, statistical trend detection techniques are applied to analyze the preprocessed data, validate the existing of software aging and quantify the effect of aging in system resources. Periodically monitored and collected data on the attributes responsible for determining the “health” of the executing software are applied to complete the experiment. The results of the experiment indicate significant availability for on-line detection of software aging.

Key words: Software aging; Wavelet analysis; Statistics analysis