 |
z/VM support of z/OS Guests
|
|
Updated: 03 May 2007
This page provides reference information and links to
resources you may find useful when running z/OS as a guest of
VM.
Publications and Redbooks
-
z/VM: Running Guest Operating Systems
This book is intended to help you to plan for and to operate guest operating
systems under z/VM®. It also includes sample execs to help you automate certain
tasks.
-
Using z/VM for Test and Development Environments: A Roundup
This IBM Redbook shows the strengths of z/VM and how you can use these
strengths to create a highly flexible test and production environment.
Some of the strengths of z/VM that are shown in this book are that you can run
Linux on z/VM, you can run a sysplex under z/VM, and you can develop code under
z/VM for z/TPF. You can also provision Linux guests under z/VM. A vswitch
allows you to connect all of your guests (all operating systems that run under
z/VM) easily to the network. You can simulate your production environment on a
sysplex.
Example of VM/ESA Support of S/390 Parallel Sysplex
z/VM Storage Management
VM FAQ's
Related links:
Support considerations for running z/OS on z/VM
This section contains:
-
Maximum of 24,576 virtual devices
z/VM limits an individual guest to having 24,576 (24K) virtual devices.
The architecture imposes a limit of 65,536 (64K) devices in a subchannel
set and allows up to four subchannel sets.
The limit of 24,576 is an arbitrary one that was established during the
development of VM/ESA. Each virtual device has an associated control
block (a VDEV) that consumes real memory resources. At present, one page
can hold 13 VDEVs, so 24,576 VDEVs would consume 1890.5 pages or 7.4
megabytes of memory.
Most guests never even approach this limit. However, if you wanted to
create a duplicate of a large production MVS system using z/VM, you might
require more virtual devices than the limit allows. This would prevent
you from creating a system with all the required devices. While this
limit has not, to IBM's knowledge, been a problem for customers, a
requirement with business justification for removing it would be
appropriate if it does become an issue.
Circumventing this limitation would be relatively trivial. The remedy
would probably take the form of a guest-specific virtual device limit,
designated via the User Directory, and perhaps a default limit established
in the System Configuration File. However, in the absence of a customer
requirement, there is no reason to pursue these enhancements.
-
Dynamic I/O Configuration devices
Dynamic I/O configuration is a mechanism that allows an operating system
to dynamically add, change, or remove I/O resources from its I/O
configuration. Channel paths, control units, and devices can be managed
in this way. In the z/OS environment, this facility is provided by the
Hardware Configuration Definition (HCD) tool. A graphical user interface,
Hardware Configuration Management (HCM), runs on a workstation and
simplifies the use of HCD.
HCD is also supported in the
z/VM environment and can be accessed via HCM. In addition, z/VM has
built-in commands to enable I/O configuration management
without HCD and HCM.
z/VM does not allow a guest to use the Dynamic I/O Configuration
interfaces to manage its virtual I/O configuration. This is because
adding this support would not add significantly to the capabilities of
a guest z/OS system since dynamic changes to the I/O configuration are
not used very extensively. Rather, they are usually used in response to
physical changes to the configuration, which occur over relatively long periods of time rather than minute to minute. As well, other VM facilities
provided through its command interface allow devices
to be added, changed, and removed.
Aside from testing HCD and HCM, there would be little value in supporting
dynamic I/O configuration. As such, this should not present an issue to
customers.
There are no plans to virtualize dynamic I/O configuration capabilities
and no intentions to reconsider this decision unless there is a
significant change in the breadth of the technology's applicability.
&npsp;
-
Dynamic Storage Reconfiguration
Dynamic Storage Reconfiguration is a mechanism that allows an operating
system to dynamically add memory resources to an remove them from its
configuration. Central and expanded storage can each be defined with
initial and reserved allocations. After a manual action (e.g.,
deactivating another logical partition), some or all of the reserved
memory amounts may become available and can be varied online
programmatically. Subsequently, these storage areas may be varied offline.
z/VM does not allow a guest to use the Dynamic Storage Reconfiguration
interfaces to manage its virtual memory configuration. While a virtual
machine can dynamically redefine its central storage size, this causes a
virtual system reset (i.e., requires an IPL). Expanded storage can be
dedicated to a guest, but its size cannot be adjusted dynamically. As
well, z/VM does not exploit DSR for its own use.
Guest support for DSR would enable testing guest support of the
associated interfaces, operator training in its use, and understanding
the potential benefits of exploiting it. These rather modest advantages
do not seem significant for most customers and should not be serious
issues for most z/OS customers.
z/VM exploitation of DSR might have more potential benefit, though not
perhaps to z/OS customers. Rather, in an environment that cannot sustain
a z/VM outage in order to resize a logical partition, the ability to add
central and expanded storage resources dynamically could offer
significant value. A prototype of this support has been developed and
allows z/VM's central storage size to be increased (there are some
architectural issues related to DSR for Expanded Storage). Storage
removal support was not part of the prototype because limiting the use
of dynamic storage so that it would be feasible to give it up
would have made it insufficiently useful. As a side-effect of this
development, guest support was also provided. There are currently no
firm plans to deliver this support.
-
Parallel Access Volume (PAV) Support
PAV support provides a mechanism for issuing multiple I/O operations
concurrently to a DASD volume, enabling I/O throughput to be increased.
In a DASD subsystem, one or more alias volumes can be associated with a
base volume to enable multiple I/O operations to be started concurrently.
While certain operations require the use of the base volume, most can
be started using either the base or one of its aliases.
VM has supported dedicated use of PAV for many years. This support is
explained in detail at the
Initial PAV Support page.
However, PAV was not supported for minidisks and required the guest
operating system to be PAV-aware in order to obtain the benefits of
increased throughput. In May, 2006, z/VM PAV support was extended to
eliminate these shortcomings. With the availability of the PTF for
APAR VM63952, PAV is now also supported for minidisks, allowing
multiple z/OS guests to have virtual alias devices associated with
virtual base volumes. In addition, PAV is exploited by z/VM for some of
its own I/O operations and for those of its guests, whether or not those
guests are PAV-aware.
A comprehensive description of what is supported is provided at
z/VM PAV Support page.
z/VM does not support dynamic activation of PAVs in response to I/O
demand. Rather, it requires a fixed assignment of aliases to bases.
This may introduce some operational complexity and may require
additional effort to ensure that there are
sufficient aliases assigned to each base to support the anticipated I/O
load. In some situations, this might not be possible.
Removing
the limitations of enhanced PAV support is planned for delivery in
z/VM 5.3 with exploitation of HyperPAV.
-
Specialty Engine Support
System z specialty engines provide support for designated workloads and
are less expensive than general-purpose CPs. Integrated Facility for
Linux (IFL) engines are intended to run Linux and z/VM systems with
Linux guests.
IBM System z Application Assist Processors (zAAPs) are designed to
provide a cost-effective execution environment for Java applications
under the control of the IBM Java Virtual Machine (JVM) on z/OS.
IBM System z9 Integrated Information Processors (zIIPs) are the latest
specialty processors, designed to help improve resource optimization and
lower the cost for eligible workloads. This includes certain DB2
processing, enhancing the role of the mainframe as the data hub of the
enterprise. z/OS and z/OS.e exploit zIIPs to offload software
system overhead from standard Central Processors.
z/VM V5.3
provides guest support for specialty engines.
Virtual zIIPs, zAAPs, and IFLs may be defined as part of guest
configurations and are implemented
using one of the following methods:
-
Simulation, where z/VM dispatches specialty engines for guest
virtual machines on real Central Processors (CPs).
-
Virtualization, where z/VM dispatches specialty engines for guest
virtual machines on the real specialty engines that match their type, if
those specialty engines are available in the real z/VM processor
configuration.
The virtual specialty engine types of zIIP, zAAP, and IFL are
externalized by several z/VM commands, including DEFINE CPU,
QUERY VIRTUAL CPUS, and INDICATE USER.
Guest support for specialty engines enables testing the associated Java
and data base applications that exploit them and evaluating the amount
of benefit to be expected from installing these engines in the real
environment. Virtualization support for these engines also enables
additional resources to be used by z/VM guests without increasing
software license charges. While modeling tools can be used to estimate
benefits, guest support provides significant additional value for z/OS
customers by allowing the actual workloads to be executed and the offload
benefits to be measured.
As the use of specialty engines continues to expand, z/VM guest support
is poised to provide a platform on which customers can explore their
benefits. IBM recently announced an enhancement to the z/OS
Communications Server to move a portion of its IPSec processing to zIIPs.
IBM also issued a statement of direction to enable the XML System Service
element of z/OS to take advantage of zIIPs and zAAPs. z/VM will support
these changes for z/OS guests as soon as they are available.
More details about the z/VM support for specialty engines will
be available
in SC24-6115-02, z/VM V5R3.0 Running Guest Operating Systems.
(at z/VM V5.3 General Availability).
|
|
|
|
|
 |
|
Easy ways to get answers you need about z/OS on z/VM
|
|
|
|
|
|
