Chapter 1. Release summary
3.1 Large volume support
The IBM TotalStorage Enterprise Storage Server (ESS) initially supported custom volumes of up to 10017 cylinders, the size of the largest standard volume, the 3390 model 9. This was the limit set by the operating system software.The IBM TotalStorage ESSLarge Volume Support (LVS) enhancement, announced in November 2001, has now increased the upper limit to 32760 (x’7FF8’) cylinders, approximately 27.8 GB. The enhancement is provided as a combination of IBM TotalStorage ESS licensed internal code (LIC) changes and system software changes, available for z/OS, OS/390, and z/VM.
Large volumes can only be defined on the IBM TotalStorage Enterprise Storage Server (ESS) or equivalent vendor storage subsystems.
The problem of concurrent, non-related requests has been addressed by the introduction of Parallel Access Volumes (PAVs) for the IBM TotalStorage ESS, which enables more than one I/O operation to be active on a volume.
For DFSMS/MVS components, this support is provided as a Small Programming Enhancement (SPE) on OS/390 Release 10 (HDZ11F0) and integrated into z/OS Release 1.3 (HDZ11G0).
You need to install large volume support on your systems before defining large volumes on the IBM TotalStorage ESS. Large volume support needs to be installed on all systems in a sysplex prior to sharing data sets on large volumes. Shared system/application data sets cannot be placed on large volumes until all system images in a Sysplex have the large volume support installed.
Installation of PTFs on some components will require a system IPL to activate. Check PSP bucket information for required PTFs. Check with your OEM software product vendors for changes to their products which may be required in support of large volumes.
3.1.1 Large volume design considerations
Here are some considerations for large volume design:
The S/390 I/O processor has an architectural device size limit of 32765 (x’7FFD’) cylinders.
The current software limit is 10017 cylinders.
Records in sequential data sets are located using two byte relative track addresses (TTR), which imposes a limit of 64K tracks per data set.
Control blocks such as the Data Extent Block (DEB) and channel commands such as SEEK use two bytes for cylinder and head addressing (CCHH) which imposes a limit of 64K cylinders and tracks
Chapter 3. DFSMSdfp enhancements 41
3.1.2 3390-9 overview
The 3390-9, with 10017 cylinders, gave a threefold increase in capacity over the 3390-3 and had the following advantages:
More DASD space could be made available to configurations that were approaching the limit of 64K devices that can be defined for a z/OS image. Storage management costs were reduced by having smaller configurations to
define and manage.
Infrequently used data could be kept online.
3.1.3 Limitations of the 3390-9 solution
These are some limitations of the 3390-9 solution:
With the original 3390-9, there were performance problems due to head movement; this was addressed, to a degree, by large control unit caches. The large capacity also meant the possibility of a large number of concurrent
requests which were single threaded due to the design of OS/390. Therefore they were of limited use if performance was a primary concern. This is also the case with the newer Redundant Array of Independent Disks (RAID) based subsystems, which emulate the 3390 device geometry.
3.1.4 Coexistence support
For DFSMS/MVS 1.4 and 1.5, a coexistence PTF is provided that will allow these system levels to coexist in the same Sysplex with LVS systems. You must install this PTF in order to prevent unpredictable results that may arise from systems without large volume support accessing volumes that have more than 10017 cylinders. The coexistence PTF will:
Prevent a device with more than 10017 cylinders from being varied online to the system.
Prevent a device from coming online during an IPL if it is configured with more than 10017 cylinder.
Coexistence PTFs will also be required by DFSMShsm on all releases prior to OS/390 R10 because of the updates being made to the record format in the DFSMShsm control data sets.
Coexistence support will not be available for DFSMS 2.10 and higher. Install the large volume support prior to using data on large volumes at these release levels. No coexistence support will be available for DFSMS/MVS 1.3 (unsupported release), or earlier.
3.1.5 EXCP considerations
Prior to this release, the flag used to indicate that a data set cannot be processed by EXCP, for example PDSE or extended format, is the high order bit of field DEBSTRCC, this has been moved to the high order bit of DEBSTRHH+1. This information was not documented in any of the DFSMS publications available at the time of writing this book.
3.1.6 Interfaces and vendor code
All vendors should be contacted to determine if any code changes are required to support these volumes, particularly Sort and Database suppliers.
The change in the DEB, described in the previous section, is most likely to affect SORTs, but locally written programs should also be checked to see if this change is relevant.
Products which use TTR notation as an absolute value and not relative to the start of the data set, most likely to be database related, cannot have an ending cylinder address higher than 4369. This is not new, as it was a problem for 3390-9 users. However, JES2 is an example of a subsystem which has changed its code to use a relative displacement; APAR OW49373 gives the details.
3.1.7 Performance
We did not have the opportunity to do any performance testing, but the use of PAVs will be essential if performance is a consideration for the data on large volumes. If you are running on an IBM D/T2064 processor, or equivalent, the Workload Manager (WLM) controlled Dynamic CHPID management should be evaluated as it could improve overall DASD subsystem throughput.
3.1.8 Implementation considerations
All systems sharing DASD should be capable of recognizing large volumes before they are configured on the ESS.
The LVS support will be rolled back to OS/390 V2R10; earlier systems will require PTFs to prevent large volumes from coming online during IPL or VARY online processing.
As always, disaster recovery (DR) must be considered. If volume backups are being taken for this purpose, then the DR site must be capable of providing large volumes.
Chapter 3. DFSMSdfp enhancements 43
Do not run the standard volume initialization job. With potentially over 32000 cylinders available for allocation, some thought needs to be put into sizing the VTOC, indexed VTOC, and VVDS.
3.1.9 Required support
The PTFs required for support or tolerance of large volumes by earlier releases are listed in z/OS V1R3.0 DFSMS Migration, GC26-7398. Full support for large volumes is provided in OS/390 2.10 and all levels of z/OS. For other supported levels of OS/390, toleration support is provided. This support prevents you from varying large volumes online to these systems.