z/VM host configuration

The z/VM® host system was configured to exploit the set of processors and the amount of memory that were configured as initial for the hosting LPAR.

If required by a particular test execution, these resource sets were modified during the test execution setup phase by activating or deactivating processors, and/or by claiming additional memory from the LPAR's reserved memory, as follows:

• CPUs were activated/deactivated as required, by means of the vary on/off processor <nn> command.
• Additional memory was made available by means of the set storage command.

This approach makes it possible to dynamically control the amount of CPUs and memory used by the z/VM host for a particular test execution, without having to change the activation profile defined in the hardware management console (HMC) for the LPAR.

For z/VM 6.3, the documentation does not recommend the use of expanded storage. Thus in this case all memory required for a test case was defined as central storage, and no expanded storage was defined.

For z/VM 6.2, the documentation recommends the use of expanded storage. Thus in this case 2 GiB of the memory required for a test case were defined as expanded storage, while the remaining memory was defined as central storage.

Two variants were configured for the z/VM paging devices:

• EDEV-SCSI variant:

  The EDEV-SCSI variant was composed of 16 EDEV paging devices based on SCSI devices configured in the storage server. Each SCSI device had a size of 32 GiB, resulting in a combined paging area of approximately 512 GiB. The value of the ATTRIBUTE parameter on the z/VM EDEV configuration statements was set to 2107. This is the recommended value for SCSI disks residing on DS8000 like storage servers.

  When using EDEV-SCSI devices for paging, z/VM is able to issue overlapping I/O operations for each EDEV-SCSI device (for details, see z/VM 6.3 Performance Considerations). In other words, I/O operations for a particular EDEV-SCSI paging device are parallelized.

• ECKD™ variant:

  The ECKD variant was composed of 25 ECKD 3390 devices that were configured on the storage server. Each 3390 device had a size of 30051 cylinders. The format established by the z/VM format utility CPFMTXA established 180 pages per 3390 cylinder, resulting in 20.6 GiB per device. All 25 devices composed a paging area of approximately 516 GiB.

  z/VM (as of z/VM 6.3) does not support PAV (or HyperPAV) for paging volumes. This means that only one I/O operation at a time can be in progress for a particular paging device.

On the storage server, the SCSI and the ECKD disks were organized in striped storage pools from both internal servers.

Sufficient spool space was configured in the form of ECKD devices located on the same storage server.

An additional large number of 3390 model 9 and model 27 DASDs were configured on the storage server. These DASDs were assigned to the virtual systems running Linux™ in the form of full-pack minidisks, in order to serve as Linux root and paging devices. Furthermore, relating to the 3390 DASDs, ranges of HyperPAV devices were also configured on the storage server. These HyperPAV devices were directly attached to z/VM, in turn enabling z/VM to support virtual HyperPAV devices for the full-pack minidisks assigned to the virtual systems.

The z/VM system was connected to an internal TCP/IP network via an OSA Express2 adapter. A z/VM Virtual Switch (VSWITCH) was configured connecting to that OSA adapter, enabling the virtual systems running Linux to use the VSWITCH for external network access.

z/VM monitor event data collection and z/VM monitor sample data collection were selectively enabled for the processor, storage, user, I/O, and network domains.