Discussions
Categories
- 382.6K All Categories
- 2.1K Data
- 213 Big Data Appliance
- 1.9K Data Science
- 448.4K Databases
- 221.1K General Database Discussions
- 3.7K Java and JavaScript in the Database
- 25 Multilingual Engine
- 541 MySQL Community Space
- 469 NoSQL Database
- 7.8K Oracle Database Express Edition (XE)
- 2.9K ORDS, SODA & JSON in the Database
- 500 SQLcl
- 3.9K SQL Developer Data Modeler
- 186.3K SQL & PL/SQL
- 21.1K SQL Developer
- 293.9K Development
- 9 Developer Projects
- 130 Programming Languages
- 290.6K Development Tools
- 96 DevOps
- 3K QA/Testing
- 645.6K Java
- 24 Java Learning Subscription
- 36.9K Database Connectivity
- 151 Java Community Process
- 104 Java 25
- 22.1K Java APIs
- 137.9K Java Development Tools
- 165.3K Java EE (Java Enterprise Edition)
- 17 Java Essentials
- 146 Java 8 Questions
- 85.9K Java Programming
- 79 Java Puzzle Ball
- 65.1K New To Java
- 1.7K Training / Learning / Certification
- 13.8K Java HotSpot Virtual Machine
- 94.2K Java SE
- 13.8K Java Security
- 201 Java User Groups
- 24 JavaScript - Nashorn
- Programs
- 296 LiveLabs
- 36 Workshops
- 10.2K Software
- 6.7K Berkeley DB Family
- 3.5K JHeadstart
- 5.8K Other Languages
- 2.3K Chinese
- 168 Deutsche Oracle Community
- 1.2K Español
- 1.9K Japanese
- 234 Portuguese
How to measure memory usage?
843798
Member Posts: 24,864
Hi all!
Is there any more sophisticated way of getting the memory requirements of objects at run-time?
Currently I'm using Runtime.totalMemory()-Runtime.freeMemory() to estimate the memory usage before and after loading the object in question. The difference should be the memory used by the object. I also call upon Runtime.gc() to make sure that the garbage collector is run before each of the measurements. Unfortunately Java doesn't guarantee that the garbage collector is really run and that all unused objects are removed from memory even if it is run.
Although this method gives you a rough indication of memory usage (at least for very big objets), I would like to have something more precise (and repeatable). Using the method above on two big objects created in different order leads to different results.
mneumi
Is there any more sophisticated way of getting the memory requirements of objects at run-time?
Currently I'm using Runtime.totalMemory()-Runtime.freeMemory() to estimate the memory usage before and after loading the object in question. The difference should be the memory used by the object. I also call upon Runtime.gc() to make sure that the garbage collector is run before each of the measurements. Unfortunately Java doesn't guarantee that the garbage collector is really run and that all unused objects are removed from memory even if it is run.
Although this method gives you a rough indication of memory usage (at least for very big objets), I would like to have something more precise (and repeatable). Using the method above on two big objects created in different order leads to different results.
mneumi
Comments
-
The is no guarantee how much, when or why the VM allocates and/or frees memory. In some cases, the VM might reserve memory as 'used' even though the objects occupying that memory have been freed. The only real metric is to go through by hand and calculate the size. freeMemory() is really only good for getting a general idea of resource usage.
-
But how can you measure the memory usage by hand? Java isn't very helpful for this. You can have a educated guess about the size of the basic types, but in the end you're guaranteed nothing. And for non trivial object types the things become much harder. The only way to work out the size of a complex structure I found so far was to have a big array full of these objects and use the method I have described above to measure the total memory usage.
This discussion has been closed.