z/VM New Release Considerations Review ToolAnnouncement
Each release has a number of changes in it, many of which bring additional value. z/VM 6.4 is no exception. To help customers identify the value and possible changes, IBM is providing an as-is utility called VMREVIEW which can be used to help assess the value of z/VM 6.4 and some key considerations. It is not a replacement for the PSP buckets or the official install guide. Also note that it identifies changes of interest, it does not migrate or change the actual system.
The utility has been written and tested for the following z/VM releases: 5.4, 6.1, 6.2, and 6.3. It does require a z/VM userid with Class B privileges. The VMREVIEW Utility can be found on the z/VM download page.
Notable Changes Identified by VMREVIEW:
Documented here are some of the considerations or changes that the VMREVIEW utility might highlight.
- Required Hardware - z/VM 6.4 introduces a new architecture level set (ALS) of a z196 or z114 or newer. That is, z/VM 6.4 drops support for the z10 BC and EC machines. The supported processors are: z13, z13s, LinuxONE Emperor, LinuxONE Rockhopper, IBM zEnterprise (R) EC12, IBM zEnterprise BC12, IBM zEnterprise 196, IBM zEnterprise 114
- No Expanded Storage Support -
z/VM 6.4 fulfills a previous statement of direction to drop support for expanded storage. With changes to memory
management in z/VM 6.3, expanded storage was no longer recommended, though it was tolerated. However, in z/VM 6.4, it
will not be brought online and will not be used at all. Therefore it is important to configure any expanded storage
as central storage prior to bringing up z/VM 6.4.
Additional Information on memory configuration.
Performance Information on z/VM 6.3 changesn.
- Share Setting Calculations - The scheduler has been enhanced to improve the accuracy and fairness of resource distribution based on the share settings. In the process of these changes, the share normalization process was changed. A consequence is it now adheres better to a principle of least astonishment. Previously the normalizaiton of relative share values was based on the sum of the share values for virtual machines in the dispatch and eligible lists, which is a subset of all virtual machines as it does not include those in the dormant list, that is the truly idle virtual machines. In z/VM 6.4, the scheduler now considers all virtual machines, including those in the dormant list. (Many people mistakenly thought this is how it worked in previous releases.) So for systems where there are many virtual machines in the dormant list, there is the potential for different behavior in z/VM 6.4.
- HyperPAV Support for z/VM Paging - While there has been support for user volumes to use HyperPAV for several releases, z/VM 6.4 introduces the ability for HyperPAV to be used for z/VM system I/O, in particular paging volumes. This allows multiple paging I/Os to be started at the same time to a paging volume. In the past, smaller paging volumes were often used to address the limitation of a single paging I/O at a time. This could result on many more paging volumes being used which makes administration more difficult. With z/VM 6.4, it's worth considering converting small paging volumes such as 3390-3s to larger volumes and using HyperPAV. It is still recommended to have at least as many paging volumes as logical processors if one wants to maximize paging bandwidth.
- New SCSI Management Queries for EDEVs - Configuring and managing emulated devices (EDEVs) for FCP SCSI can be confusing, especially with a lack of detailed information. z/VM 6.4 introduces new information on QUERY EDEV command and a new utility, IOEXPLOR, which can provide various information that previously unavailable, such as the physical serial number. Customers with EDEVs will want to be aware of this.
- No Install via Tape - Tape drives are no longer supported for install or service in z/VM 6.4, only DVD and electronic.
- Dynamic SMT support - Simultaneously Multithreading support was introduced via service to z/VM 6.3. However, to enable or disable SMT, one had to change a statement in the system configuraiton file and reIPL z/VM. In z/VM 6.4, we introduce the ability to dynamically vary the number of active threads per core. The system must still have a configuration statement enabling, or "opting" in to, SMT, though it can start with a single thread per core, also known as SMT-1. The performance of a non-SMT system and SMT-1 system are effectively equivalent.
- VSwitch reset counters function - The CP QUERY VSWITCH command can be used to get information on Virtual Switches, including information on discarded packets or errors. This is helpful for identifying problems. However, previously, these counters accumulate for the life of the VSwitch. So error counters could continue to show a non-zero value months after the condition. z/VM 6.4 allows these counters to be cleared or reset to facilitate tracking of problems.
- Support for 2TB of Memory - A major highlight of z/VM 6.4 is the amount of supported memory has doubled from 1 TB to 2 TB. For those systems that are memory constrained or growing, this can be a huge benefit. The limit for individual virtual machines remains at 1 TB. If you do explore the greater memory capability please do ensure you have sufficient page and dump space.
- Reconfiguring XSTORE as a real memory resource
- Tuning your system to reduce eligible list usage
- Enabling SMT in z/VM