Test metrics

Workload performance metrics

All workload performance metrics are not presented as absolute values, but as the relative values with respect to the reference result.

Database BI workload relative throughput: relative throughput of the Database BI workload with respect to the reference result. For details, see Database BI workload.

File system I/O workload relative throughput: relative throughput of the File system I/O workload with respect to the reference result. For details see File system I/O workload.

Java™ workload relative throughput: relative throughput of the Java workload with respect to the reference result. For details, see Java workload.

Network workload relative throughput: relative throughput of the Network workload with respect to the reference result. For details, see Network workload.

Transactional WAS workload relative throughput: relative throughput of the transactional WAS workload throughput with respect to the reference result. For details, see Transactional WAS workload.

Overall workload relative throughput: combined relative throughput of all workloads. This gross metric is the sum of the relative throughput of each of the individual workload variants, divided by the number of workload variants.

z/VM performance metrics

When applicable, the z/VM metrics are ascribed to values presented in z/VM Performance Toolkit screens. In this case, the origin of a metric is listed by naming the z/VM Performance Toolkit report identification, and the name of the origin item in the report. Occasionally, multiple reports or items are listed if more than one context (such as general versus virtual machine specific) applies. For details about the z/VM Performance Toolkit, see z/VM 6.3 Performance Toolkit Reference.

For better comparability, the preferred units used for metrics in this documents are

• based on bytes for memory,
• based on bytes/s for memory rates,
• based on processors for CPU loads.

If necessary, values presented in different units by the z/VM Performance Toolkit were converted to these units.

CP attributed to guest CPU load: the CPU load spent for processing z/VM control program code that relates to individual virtual systems, such as simulating the execution of a privileged instruction for a virtual system. CP attributed to guest CPU load is determined by subtracting the emulation CPU load from the CPU load.

CP system CPU load: the CPU load spent for processing z/VM control program code that cannot be attributed to individual virtual systems. The CP system CPU load for the whole z/VM system is reported in [FCX100:%SYS] (separately for each processor and processor type).

CPU load: the processing power consumed by one or more virtual systems, measured in fractions of real processors. Each virtual processor can consume a maximum of one real processor. The CPU load for the whole z/VM system is reported in [FCX100:%CPU] (separately for each processor and processor type). The CPU load of individual virtual systems is reported in [FCX112:%CPU].

DPA size: the size of the z/VM dynamic paging area (DPA). The DPA is the region of real memory that z/VM maintains for resident user pages. It accounts for the by far largest part of real memory used by z/VM. The DPA size is reported in [FCX103:Total DPA size].

emulation CPU load: the CPU load spent for processing virtual system workload, that is, while program code directly originating from virtual systems is executed. The emulation CPU load is reported for the whole z/VM system in [FCX100:%EMU] (separately for each processor).

instantiated memory: the amount of memory in use by a virtual system. It is composed as follows:

• memory accessed (read or written) by code running in the virtual system
• minus memory that was released back to CP, and not accessed again

The instantiated memory is usually smaller than the sum of resident memory and stored memory, because it counts memory pages kept in both places as one page. The instantiated memory of individual virtual systems is reported in [FCX292:Inst].

page rate: the sum of the page read rate and the page write rate. May be shown for one or more virtual systems, or for the whole z/VM system.

page read rate: the amount of data read per second from the z/VM paging store in MiB/s. May be shown for one or more virtual systems, or for the whole z/VM system. The overall z/VM page read rate is reported in [FCX103:Page read rate]. The virtual system page read rate is reported in [FCX290:<Movement/s> Reads] for z/VM 6.3, and in [FCX113:<Page Rate> Reads] for z/VM 6.2 and 6.3.

page write rate: the amount of data written per second to the z/VM paging store. May be shown for one or more virtual systems, or for the whole z/VM system. The overall z/VM page write rate is reported in [FCX103:Page write rate]. The virtual system page read rate is reported in [FCX290:<Movement/s> Write] for z/VM 6.3, and in [FCX113:<Page Rate> Write] for z/VM 6.2 and 6.3.

real CPU count: the number of CPUs configured for z/VM. For the tests described in this paper, central processors (CPs) were configured at the LPAR level (see Hardware), but Integrated Facilities for Linux™ (IFLs) could have been used as well. The count of real CPUs is reported in [FCX304:CPU], separately for each processor type.

real memory: the amount of memory configured for z/VM. The real memory size configured for z/VM is reported in [FCX103:Total real storage].

resident memory: the amount of memory backed in the z/VM DPA for a particular virtual system. The resident memory of individual virtual systems is reported in [FCX292:T_All].

stored memory: the amount of memory stored in the z/VM paging store for a particular virtual system. The stored memory of individual virtual systems is reported in [FCX292:AUX].

virtual memory: the amount of virtualized memory defined for a virtual system. The virtual memory size of individual virtual systems is reported in [FCX292:Base Space Size].