JVMSERVER statistics
The JVM (SJ) domain collects statistics for JVM servers, including statistics on heap storage and garbage collection. Each JVM server is represented by a JVMSERVER resource.
You can get some information about the JVM server by inquiring on the JVMSERVER resource. The resource provides information such as the initial, maximum, and current heap size and the garbage collection policy that is being used by Java™. Unlike pooled JVMs, the garbage collection is handled by Java automatically depending on the policy that is specified.
The DFH0STAT and DFHSTUP statistics programs provide more in-depth information about a JVM server:
- The statistics report how long Java applications are waiting for threads in the JVM server. If the waits are high and many tasks are suspended with the JVMTHRD wait, you can increase the value of the THREADLIMIT attribute on the JVMSERVER resource to make more threads available to the applications.
- The statistics report the heap sizes of the JVM. If the heap size after garbage collection is close to the maximum heap size, garbage collection might be occurring too often and you might need to increase the maximum heap size. If the peak heap size is much lower than the maximum heap size, you can either run more work in the JVM server, or edit the JVM profile and reduce the maximum heap size to save on storage.
- The statistics report the system threads in the JVM server. System threads are used to collect statistics and are also used by inquire and browse commands, but not by applications. You can find out how many times the JVM server was accessed for information and the associated processor usage. If the number is high, you might change the statistics interval or stop the inquire and browse requests.
- The statistics report major and minor garbage collection events. Minor garbage collection is only available on certain policies, so you might want to change the policy based on the information in the statistics. For more information, see Garbage collection and heap expansion.
These statistics can be a good starting point for tuning the performance of your Java workload.