Research On Low Power Garbage Collection Of Java Memory Management

Java technology has been very widely used currently in various fields. Under the concept of green computing, energy consumption optimization technology has been the research hotspot. Energy consumption optimization of Java platform memory management has become one of the key technologies of embedded platform and Java technology platform. In particular, with reduce of CPU energy consumption and the restriction of embedded platforms, Java memory energy optimization will have an enormous role in pushing Java technology applications in the embedded field. Java platform, the system is able to implement power optimization measures according to the Specific circumstances of running, can not only effectively save memory management energy, and can meet the needs of program performance. Therefore, based on the Specific characteristics and application characteristics of the embedded platform, aiming at energy consumption issues of memory management system for Java platform, on banked-memory management architecture, this paper research and analyze low-power technology of garbage collection and implement EAMC low-power garbage collector on the RVM. Evaluation results show that these optimization techniques can indeed reduce energy consumption of memory management system on Java platform.Firstly, this paper researches the memory behavior of Java applications with detailed analysis of large numbers of typical Java application memory behavior data on the Java run-time system. Through analysis of the data, this paper finds obvious qualities of Java application's memory allocation model and memory usage tracks. So this paper focuses on qualities of the behavioral pattern of trajectory, and on the analysis and summary of Java memory behavior obtains the general laws of Java memory trajectory: the staged representation, periodicity and stationary. These qualities have a significant importance for the Java virtual machine garbage collection and Java programs energy consumption optimization. Secondly, on the basis of the rules, this paper researches the methods of predicting the behavior of garbage collection, primarily from the two sides of memory allocation and memory release, and on the basis of the two sides to predict the memory trajectory. Predicted results show that the prediction accuracy rate is meanly up to 80% or more. Finally, on the analysis and forecasting of Java memory, this paper design EAMC energy aware mark-compact garbage collector. EAMC garbage collector is mainly composed of hardware monitor module, memory trajectory forecasting module and garbage collection controller module. In addition, EAMC garbage collector changes the garbage collection trigger factor from the previous performance trigger factor into the trigger method on the combination of performance trigger factor and energy consumption trigger factor. This paper implements EAMC garbage collector on the realization on RVM and takes its experimental testing. In the experimental test of EAMC garbage collector, this paper establishes power model on the general system characteristics of embedded platform, and compare the energy consumption saving effect with traditional mark-compact garbage collector. The test data show that, compared to the traditional mark-compact garbage collector, EAMC low-power garbage collector averagely saving 10% of memory management energy consumption, while the effect of applications with slower memory allocation rate is more evident.