基于分形的软件衰退预测

Fractal-based Approach for Software Aging Prediction

许多研究人员指出导致软件性能衰退的主要原因之一是系统资源的耗尽,而软件系统实时运行时采集到的系统资源耗费数据展示出了分形这一几何特征,因此应用了基于分形分析的方法,计算反映软件衰退状态的霍尔德指数,实现软件性能衰退的预测。首先改进和扩展了霍尔德指数计算算法,使得能够计算多维霍尔德指数;然后使用经典的统计方法预测软件衰退导致宕机的时间,接着用实验验证了改进的多维霍尔德指数计算算法的精度,最后使用软件运行中采集到的系统资源耗费数据进行了仿真实验,实验结果显示该算法有效的预测宕机时间,从而为执行软件自愈策略提供依据。

A number of recent studies have reported the phenomenon of ＂software aging＂, characterized by progressive performance degradation and a sudden crash/hang of a software system due to exhaustion of operating system resources, fragmentation and accumulation of errors. Because the data of resource usage collected from software system display fractal characteristics, a fractal-based method was adopted to calculate the holder exponent of resource usage for software aging. First, the existed algorithm was improved to calculate multidimensional holder exponent and check the improved algorithm＇s accuracy; then statistical method was adopted to predict the downtime of software system; finally the proposed method was tested using the performance parameters data collected from a realistic software system to evaluate the forecasting performance.