Windows环境软件分级再生策略研究  

Research on software hierarchical regeneration strategy in Windows environment

作　　者：王艳超 姚江毅 李雄伟 WANG Yanchao;YAO Jiangyi;LI Xiongwei(Shijiazhuang Campus,Army Engineering University,Shijiazhuang 050003,China)

机构地区：[1]陆军工程大学石家庄校区,河北石家庄050003

出　　处：《现代电子技术》2023年第4期79-84,共6页Modern Electronics Technique

基　　金：国家自然科学基金项目(62071483)。

摘　　要：为降低软件系统再生成本,提高再生效益,文中在预测软件系统老化趋势的基础上,利用细粒度再生时间短和粗粒度再生效果好的特点,提出一种双粒度分级再生策略。该策略包括系统级再生和服务级再生。首先,以Windows环境下的综合信息管理系统为背景,搭建Apache Web服务器老化数据采集平台,通过修改服务器参数实现加速老化实验并采集系统老化数据;然后,将时域卷积神经网络模型(TCN)应用于老化趋势预测,对采集的老化数据进行特征提取并建模,得到该方法的预测精度高于循环神经网络;最后,提出一种再生决策函数评估平均再生成本的变化,依据再生成本的变化产生再生请求,实施分级再生。结果表明,双粒度分级再生策略平均再生成本降低7.6%,系统运行时间延长22.7%。与单纯的系统级再生策略相比,该策略能够降低平均再生成本,提高操作系统可靠性。In order to reduce the regeneration cost of the software system and improve the regeneration efficiency,based on the prediction of the aging trend of the software system,a dual-granularity classification regeneration strategy is proposed according to the characteristics of short fine-grained regeneration time and good coarse-grained regeneration effect. This strategy includes system-level regeneration and service-level regeneration. An Apache Web server aging data acquisition platform is built based on the comprehensive information management system in Windows environment. By modifying the server parameters,the accelerated aging experiment is realized and the system aging data is collected. The time domain convolution neural network(TCN) model is applied to the aging trend prediction. The feature extraction and modelling of the collected aging data are conducted to get the prediction accuracy of this method,which is higher than that of the cyclic neural network. A regeneration decision function is proposed to evaluate the change of average regeneration cost and produce regeneration request according to the change of regeneration cost,so as to implement hierarchical regeneration. The results show that the average regeneration cost of the dual-grained graded regeneration strategy is reduced by 7.6%,and the system operation time is extended by 22.7%. In comparison with the simple system-level regeneration strategy,this strategy can reduce the average regeneration cost and improve the reliability of the operating system.

关 键 词：软件分级再生 老化数据采集 老化预测 特征提取 再生决策函数 再生成本评估 再生效益分析 

分 类 号：TN911-34[电子电信—通信与信息系统] TP311.5[电子电信—信息与通信工程]