Performance for IBM Z

Parallel Access Volumes

A PAV capability represents a significant performance improvement by the storage unit over traditional I/O processing. With PAVs, your system can access a single volume from a single host with multiple concurrent requests.

You must configure both your storage unit and operating system to use PAVs. You can use the logical configuration definition to define PAV-bases, PAV-aliases, and their relationship in the storage unit hardware. This unit address relationship creates a single logical volume, allowing concurrent I/O operations.

Static PAV associates the PAV-base address and its PAV aliases in a predefined and fixed method. That is, the PAV-aliases of a PAV-base address remain unchanged. Dynamic PAV, on the other hand, dynamically associates the PAV-base address and its PAV aliases. The device number types (PAV-alias or PAV-base) must match the unit address types as defined in the storage unit hardware.

You can further enhance PAV by adding the IBM HyperPAV feature. IBM HyperPAV associates the volumes with either an alias address or a specified base logical volume number. When a host system requests IBM HyperPAV processing and the processing is enabled, aliases on the logical subsystem are placed in an IBM HyperPAV alias access state on all logical paths with a specific path group ID. IBM HyperPAV is only supported on FICON® channel paths.

PAV can improve the performance of large volumes. You get better performance with one base and two aliases on a 3390 Model 9 than from three 3390 Model 3 volumes with no PAV support. With one base, it also reduces storage management costs that are associated with maintaining large numbers of volumes. The alias provides an alternate path to the base device. For example, a 3380 or a 3390 with one alias has only one device to write to, but can use two paths.

The storage unit supports concurrent or parallel data transfer operations to or from the same volume from the same system or system image for IBM Z or S/390® hosts. PAV software support enables multiple users and jobs to simultaneously access a logical volume. Read and write operations can be accessed simultaneously to different domains. (The domain of an I/O operation is the specified extents to which the I/O operation applies.)

Multiple allegiance

With multiple allegiance, the storage unit can run concurrent, multiple requests from multiple hosts.

Traditionally, IBM storage subsystems allow only one channel program to be active to a disk volume at a time. This means that, after the subsystem accepts an I/O request for a particular unit address, this unit address appears "busy" to subsequent I/O requests. This single allegiance capability ensures that additional requesting channel programs cannot alter data that is already being accessed.

By contrast, the storage unit is capable of multiple allegiance (or the concurrent execution of multiple requests from multiple hosts). That is, the storage unit can queue and concurrently run multiple requests for the same unit address, if no extent conflict occurs. A conflict refers to either the inclusion of a Reserve request by a channel program or a Write request to an extent that is in use.

z/OS Distributed Data Backup

z/OS Distributed Data Backup (zDDB) allows hosts, which are attached through a FICON interface, to access data on fixed block (FB) volumes through a device address on FICON interfaces.

If the zDDB LIC feature key is installed and enabled and a volume group type specifies either FICON interfaces, this volume group has implicit access to all FB logical volumes that are configured in addition to all CKD volumes specified in the volume group. In addition, this optional feature enables data backup of open systems from distributed server platforms through a IBM Z host. The feature helps you manage multiple data protection environments and consolidate those into one environment that is managed by IBM Z . For more information, see z/OS Distributed Data Backup.

z/HPF extended distance

z/HPF extended distance reduces the impact that is associated with supported commands on current adapter hardware, improving FICON throughput on the I/O ports. The storage system also supports the new zHPF I/O commands for multitrack I/O operations.

zHyperLink

zHyperLink is a short distance link technology that is designed for up to 10 times lower latency than zHPF. It can speed up transaction processing and improve active log throughput. zHyperLink is intended to complement FICON technology to accelerate I/O requests that are typically used for transaction processing.