Research Of Java Object Life Behaviors

Garbage-collected languages such as Java and C# are now increasingly prevalent. However, the performance of traditional garbage collectors is suffering more and more degradations. In order to improve the performance of garbage-collected applications it must make the garbage collector cooperate with all the components in the memory hierarchy adequately to minimize the cost of the system and maximize the performance of the garbage-collected applications. So it is necessary to analyze the memory visiting behaviors based on objects to tune the traditional garbage collectors or direct the design of the new algorithm of garbage collection.First we analyze the reason why existing garbage collectors suffer low performance and the new generation garbage collectors have a high performance. At the same time we introduce the common tracing algorithms in the international research field.Second, based on the deep analysis of the principles of the famous open source Java virtual machine Jikes RVM and MMTk, we present a framework for tracing and counting objects life behaviors based on Jikes RVM. The use of this framework is to perform a complete tracing of the whole process of garbage collection so that it can achieve the capture and counting and analysis of the object life behaviors.Finally, we implement the framework and test it. Then we apply it to perform a complete tracing of the classic Java application SPECjvm98 benchmark. All the life behavior records are captured to be counted and analyzed, harvesting some rules of memory visiting to direct the design of the new algorithm of garbage collection.