基于时间序列分析的应用服务器性能衰退模型  

Performance Degradation Model of Application Server System Based on Time Series Analysis

作　　者：徐文彬[1] 齐勇[1] 侯迪[1] 刘亮 

机构地区：[1]西安交通大学电子与信息工程学院 [2]IBM中国研究中心,北京100087

出　　处：《西安交通大学学报》2007年第4期426-429,共4页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(60473098);IBM中国研究中心联合开发项目

摘　　要：针对JUFrame应用服务器老化情况,设计了多种类型的客户请求程序和服务器端程序,记录了各种参数共计5类36个.对影响应用服务器中间件性能的主要指标“内存利用率”,采用时间序列分析方法建模,经采样数据的统计分析得出了自相关函数明显拖尾、偏自相关函数明显截尾的结果.模型的仿真结果表明,预测值与原始数据拟合良好,且可作为系统再生的依据,它适用于预测实际的运行系统.与其他软件老化预测方法比较,所提模型更切合系统的实际运行状况,有利于提高系统的可用率.Many kinds of client request applications and server programs are designed for the aging of JUFrame application sever, in which 36 parameters in 5 groups are recorded. The key index ＂memory usage＂ that affects the performance of middleware of application server is modeled based on time series analysis. The results that the auto correlation function （ACF） is clearly taildragged and the partial auto correlation function is clearly tail-truncated can be drawn via detailed statistical analysis of sampled data. The result of model simulation shows that the predicted value fits the original data well and can be regarded as the basis of system rejuvenation, and is applicable for predicting runtime systems. Compared to other software aging prediction methods, the proposed method is more suitable for the practical case of systems, and the availability of the system is improved.

关 键 词：时间序列 软件老化 自相关函数 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]