IBM Professional Papers

• Japanese

Marusa, a Memory Analysis Tool for Java VM

It is important to analyze the memory footprint of JVM™ itself, as well as that of Java program, for fully breaking down the memory footprint of a Java™ program. However, there was no such tool that analyzes both the memory usage in each of JVM data structures (JVM-level information) and the process management information from OS (OS-level information). Our memory analysis tool Marusa breaks down the footprint of JVM using the both levels of the information. It can also break down the amount of physical memory usage. These features are useful for detecting and analyzing problems caused by large footprint, and verifying the effect of changes in configuration parameters.

Words & Phrases：
Marusa, Java, memory footprint analysis, analysis tool, JVM configuration

1. Introduction
   
   
2. Elements consisting of process memory
   2.1  OS-level process memory elements
   2.2  JVM-level process memory elements
   
   
3. Memory analysis in Marusa
   3.1  Our approach
   3.2  Implementation of Marusa
   
   
4. Use cases of Marusa analysis
   4.1  Analysis of a single memory snapshot
   4.2  Analysis of the variation over time
   4.3  Exploiting Marusa for optimizing the size of shared class cache
   
   
5. Conclusion

[ PROVISION | No.59 ]