Observation

The transactional database workload is depicted using the same scheme as previously used.

In Figure 1 the utilization values (dark green lower area) show the active CPU usage of the three other guest systems that comprise the mixed workload, in units of CPUs. CPU usage is based on Linux® sar data from the three guest systems. Likewise, the steal values (bright green upper area) show how much CPU power was not available in spite of a CPU demands by the guests because it was needed elsewhere within the source z/VM® system.

It can be seen that after an initialization phase around 1.8 CPUs (with spikes up to 2.5 CPUs) are consumed by the three other systems. Above the steal curve, an almost constant area reflecting unused CPU resource remains.

As the relocation process for the guest running the transactional database workload starts after about seven minutes, the steal value for the remaining guests exhibits a small increase at the expense of the utilization value can be observed. However, there is no sharp decline of the utilization value as the relocation process reaches its end (as was the case for the relocated transactional database workload).

As the relocation process is completed, the utilization value of the remaining guests sharply rises to a value of about 2.7 CPUs, while at the same time the steal values start remaining almost insignificant below 0.1 CPUs. As before the relocation, above the steal curve, an almost constant area reflecting CPU capacity not requested from these guests (idle times).