Xen虚拟机环境下的软件衰退研究

发布时间：2018-01-01 22:21

本文关键词：Xen虚拟机环境下的软件衰退研究　出处：《南京理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：Xen是一种被广泛应用的虚拟化软件平台,具备出色的隔离特性。隔离特性是通过引入VMM层实现的,Xen是VMM的一种具体的实现载体。由于本文的研究涉及到修改VMM的源代码,而Xen开放源代码的特性恰好为本文的衰退分析研究提供了实证基础,因此本文采用Xen作为虚拟化系统的研究载体。在分析了国内外软件衰退的研究现状的基础上,论文指出现有的两大研究方向分别是基于理论模型和基于测量的方式。前者的主要思想是：借助马尔科夫过程,Petri网等数学工具刻画系统运行时状态变迁的模型,并应用数学方法求解最优自愈时间间隔,适用于具有静态衰退剖面的场景。而后者的主要思想是：持续的监测系统运行时的表征性能参数,分析系统当前所处的性能状态,并综合考虑目前的实时负载等因素确定最优自愈时刻。通常利用数据挖掘和人工智能方法建模分析,适用于具有可变衰退剖面的场景。基于上述背景分析,我们首先针对Xen虚拟化系统,进行系统监测,设计并实现了一种系统监测工具,负责从VMM层采集运行时的VMM和VM资源使用状态信息,以及主要系统功能部件的活动信息；在采集数据的基础上,研究衰退分析方法,并设计了衰退分析系统,提出的衰退分析方法考虑了不同的负载特征对于衰退预测和识别准确性的影响,建立了负载模型用于区分不同负载模式,应用主成分分析方法对于资源使用信息进行深入分析,识别导致衰退的关键参数,进一步地,研究了改进的马尔科夫和人工神经网络相结合的衰退预测方法识别和预测软件衰退；结合负载模型和衰退预测方法,提出了一种自适应的衰退分析方法,并进行了系统验证。
[Abstract]:Xen is a widely used virtual software platform, has good isolation characteristics. The isolation characteristic is through the introduction of VMM layer, Xen VMM is the carrier of a concrete. Because of this research involves modifying the source code of VMM, and the characteristics of Xen open source code just provides demonstration based on the regression analysis, this paper uses Xen as the carrier of the virtual system.
Based on the analysis of the research status of domestic and foreign software recession on the paper pointed out that the two existing research directions are based on the theoretical model and measurement methods based on the former. The main idea is: with the help of Markoff, Petri and other mathematical tools to depict system runtime state transition model, and the application of mathematical methods for solving the optimal the healing time interval, suitable for static scenes. The main idea of the recession section and the latter is: characterization of performance parameter monitoring system of continuous time, the performance analysis system, and considering the real-time load current and other factors to determine the optimal self-healing moment. Usually use data mining and artificial intelligence method modeling and analysis for, with a variable profile recession scenario.
Based on the analysis of the above background, we firstly Xen virtual system, monitoring system, design and implement a system monitoring tool, responsible for the use of state information from the VMM layer collects the runtime VMM and VM resources, the main functional parts and system information; in collecting data on the basis of the analysis of recession, and design the recession analysis system, the recession analysis method considering different load characteristics for the impact of the recession prediction accuracy and recognition, a load model for different load model, using principal component analysis method to analyze the resource usage information, resulting in the identification of key parameters, decline further, study combined with the improved Markoff and artificial neural network prediction method to identify and predict the recession software recession; a combination of load model and recession prediction method, extraction An adaptive regression analysis method is presented and the system verification is carried out.

【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP302

【参考文献】