The VolanoMark 10/100 loopback test was ran to compare the message throughput using different JVM heap sizes for both the client and server JVM on the 2.4.4 + MQ Scheduler kernel.
Increasing the initial JVM heap size from 8 MB to 64 MB for both the client and server JVMs improved :
8P throughput by 27 %
4P throughput by 17 %.
The throughput improvement most likely resulted from decreasing the number of times the JVM has to do garbage collection of the heap and reducing the number of allocation failures from the heap.
Increasing the initial JVM heap size from 8 MB to 64 MB also reduced
the idle time for the 4P system.
The 8P system with the MQ scheduler and 64 MB initial heap size
tends to hang and idle time is affected.
| Heap size | UP | 4P | 8P |
| 8MB System | 42.02 | 39.99 | 33.08 |
| 8MB User | 57.50 | 45.84 | 42.52 |
| 8MB Idle | 0.48 | 14.17 | 24.40 |
| Heap size | UP | 4P | 8P |
| 64 MB System | 41.69 | 41.39 | ----- |
| 64 MB User | 57.74 | 49.73 | ----- |
| 64 MB Idle | 0.57 | 8.89 | ----- |