基于硬件性能计数器的软件异常监控模型

Software Exception Monitoring Model Based on Hardware Performance Counter

文章的目的是实现对长时间高可靠性运行软件进行自动运行监控,并及时地识别软件运行过程中出现的未知异常。基于内置于CPU中的硬件性能计数器(CPUHardwarePerformanceCounter,CHPC)所采集的性能数据,应用朴素贝叶斯模型(NaiveBayesian),提出了一个用于识别软件运行过程中未知异常的应用模型和参数学习方法。在此基础上开发了软件异常识别系统“SoftDiagnose”。数值实验结果得到以下结论:基于CHPC的朴素贝叶斯方法能够利用很少的数据识别不明原因的资源抢占、函数暂停、病毒感染等软件异常,在实验环境下识别率高达99%。

The objective of this paper is to monitor the run-time status of some key-role softwares using an automatic method.It also needs some ways to warn the system administrator in time when some software components are working on exceptional status.Based on the internal CPU Hardware Performance Counters（CHPC） and Naive Bayesian method,a software status diagnosis model is presented,with parameter learning method.A software monitoring system named ＂SoftDiagnose＂ has developed.Numeric experiment result presents the conclusion that only based on few CHPC performance data,the model can recognize unknown software exceptions under a variable environment,like resource insufficiency,virus, etc.and the correction rate about 99% under our experiment environment.