The Java Virtual Machine Regression Analysis

With the continuous development of the Java platform and the good cross-platform feature, more and more application systems, embedded systems tend to use Java application platform, which makes Java’s application area expanded. Although the Java virtual machine has many advantages, the application service running on it is still often observed with an aging phenomenon. Few studies focus on the Java virtual machine which contributes to the aging of application services, instead focuses on the application service layer and the operating system layer. On the other hand, further studies of the Java virtual machine can portray itself more accurately which can identify the reasons as well.For the above purposes, based on the structure of Java virtual machine this paper first definite the performance metric indicator system for the analysis of the Java virtual machine, which includes all Java virtual machine’s units:system service unit, memory management unit, execution unit and virtual layer of operating system unit. This metric system can describe the whole running state of the Java virtual machine. Next, this paper has designed and implemented the Java virtual machine performance capture system. The system is designed by the idea of Agent monitor and can collect every component in parallel. The collected data possess merits such as time granularity controllability, integrity and etc. Also the system has less impact on the running Java virtual machine. Finally, this paper analyses the collected performance indicators by statistical method, using principal component analysis method to remove the relation between the inner indicators, using multiple linear regression method to build the aging model. The experimental results show that the main reasons for the aging of the Java virtual machine are system service unit and memory management unit, in a further analysis, finding memory management component and code cache component are the key memory management factor to lead Java virtual machine aging.