Software Xen Virtual Machine Environment Recession Research

Xen is a widely used virtualization software platform with a good isolation characteristic. The isolation characteristic is implemented through the introduction of a middle-layer software named VMM, Xen is a specific realization of the VMM. Since this research involves modifying the source code of VMM, the characteristics of Xen open source code just provide empirical basis to this analysis, so this paper uses Xen virtualization system as the research carrier.Based on the analysis of the research situations of domestic and foreign in the filed of software aging, we conclude two aging analysis methods:the theoretical model based method and the measurement based one. The former’s main idea is to characterize the system operation state transition model through mathematical tools, such as Markov process and petri nets, and apply these mathematical methods to solve the optimal rejurvenation time interval. Generally, this method can be applicable to the scenarios with a static aging profile. The latter’s main idea is to continuously monitor runtime system performance parameters, analyze the current system state, and then determine the optimal time to rejurvenation by taking into system workloads and other factors consideration. Generally, the date mining technology and the artificial intelligence modeling technology can be used to analyze software aging, which is more suitable to scenarios with a varying aging profile.We first monitor a virtualization system with Xen, design and implement a system monitoring tool responsible for VMM and VM resource usage information collection at runtime from the VMM layer, as well as the activities information of the main system components; Using the collected data, we study a software aging analysis method and design a software aging analysis system. The proposed software aging analysis method takes the impact of workload characteristics on prediction accuracy into consideration, builds workload models to recognize different workload patterns. According to resource usage data in a specific workload model, the principal component analysis method is used to identify crucial aging indictors and then an improved aging prediction algorithm using a combinatory of the markov model and artifical intelligence network is proposed and applied to these key aging indictors to predict and verficate software aging. At last, an adaptive software aging analysis method is put forward and verificated, which applies a combinatory of the workload models and the aging prediction alogorithm.