WebSphere® Real Time for AIX® on 64-bit POWER® is optimized for consistently short GC pauses rather than the highest throughput performance or smallest memory footprint.
Certified systems have sufficient clock granularity and processor speed to support WebSphere Real Time for AIX on 64-bit POWER performance goals. For example, a well written application running on a system that is not overloaded, and with an adequate heap size, would normally experience GC pause times of about 3 milliseconds, and no more than 3.2 milliseconds. During GC cycles, an application with default environment settings is not paused for more than 30% of elapsed time during any sliding 60 millisecond window. The collective time spent in GC pauses over any 60 millisecond period should add up to about 18 milliseconds.
The main sources of variability in a standard JVM are garbage collection pauses. In WebSphere Real Time for AIX on 64-bit POWER, the potentially long pauses from standard Garbage Collector modes are avoided by using the Metronome Garbage Collector. See Using the Metronome Garbage Collector.
Class data sharing provides a transparent method of reducing memory footprint and improving JVM start time. To learn more on class data sharing see Class data sharing between JVMs
From WebSphere Real Time for AIX on 64-bit POWER V2 SR3, the 64-bit JVM uses compressed references. When using compressed references, the JVM stores all references to objects, classes, threads, and monitors as 32-bit values. Using compressed references improves the performance of many applications because objects are smaller, resulting in less frequent garbage collection and improved memory cache utilization. For further information about compressed references, see the Memory Management section of the Diagnostics Guide.
From WebSphere Real Time for AIX on 64-bit POWER V2 SR3, regular Java™ threads can run with the policy SCHED_RR in addition to the default policy SCHED_OTHER. When running with the policy SCHED_RR, threads can run with a Linux® priority 1 - 10, giving you finer control over your application. For more information about thread scheduling and dispatching, see Thread scheduling and dispatching.