Test workload

A Java™ workload was executed within a Linux™ on system z operating system member running within the z/VM® guest virtual machine that was relocated.

The workload was customized for producing a constant pressure on the CPU and memory resources such that the z/VM system was loaded up to but not beyond its respective capacities. This constant workload provides for the comparability of the relocation operations.

The Java setup parameters are listed in Table 1.

The heap size was determined such that a maximum amount of the memory available within the Linux guest operating system was claimed by the workload.

The optthruput garbage collection algorithm was selected because it causes intensive memory usage during the garbage collection phases. This was intended to produce a memory usage pattern with a large amount of updated pages spread over the whole Java heap size.

The execution time of 600s was determined such that on the source z/VM member the workload produced a stable resource consumption after an initialization phase, and, after the live guest relocation, on the target z/VM member again resumed a stable resource consumption.

Aside from this Linux guest virtual machine, no other significant workloads were performed within the z/VM systems.