Yes your reading and interpretation is essentially correct.

VSAM was implemented at the OS level with DOS/VS (and IIRC, DOS/VS was first to have VSAM with MVS|VS1 to follow), where with OS (OS/VS1 or MVS) it was implemented at the address space level.

Whoever did your migration, if they did not have a background involving DOS/VS_ and just did a flat migration to MVS (z/OS), you can get royally shafted.

The SHARE OPTIONS between the two systems are very different and one has to know and understand this to do a proper migration and Catalog structures are very different between the two systems. Where you would do backups by CATALOG, a CATALOG does not OWN the volumes in an MVS shop. But they did in a DOS/VS shop.

And I hate to break this to you at this late date, but if the migrators didn't know it, the z/VSE system was an XA I/O system and so the performance increase for I/O that one expected in days of yore in going to z/OS will not be there (until DOS/VSE/ESA, DOS systems were BASE S/370 using the OLD SIO/SIOF, etc. and not SSCH and related instructions).

You may actually lose performance in the z/OS environment as a result.

Regards,
Steve Thompson


On 09/04/2018 07:42 PM, Tony Thigpen wrote:
My main background is z/VSE but now I have to manage a bunch of z/OS sites, including one that recently converted from z/VSE to z/OS.

On z/VSE, share-option 4 means that VSAM will prevent any read or write integrity exposures when multiple tasks are accessing the same VSAM file.

z/VSE VSAM will internally lock any CI that is being updated so that nobody else can update the CI. This ENQ/DEQ is handled by the IBM provided VSAM IO routines at the task level. Additionally, VSAM will flush all update buffers after a write or update. And, it will not buffer reads when reading a share-option 4 file. (I am being somewhat general in the descriptions, so the details are a little more complicated.) All this to make sure that the records on disk and the records in buffers match.

Now, with z/OS, my reading of the VSAM Demystified RedBook leads me to the following: 1) Share-option 4 allows multiple open for update, but expects the program, not the VSAM subsystem, to perform the ENQ/DEQs. 2) If a program does not perform ENQ/DEQs, then data integrity is lost as multiple tasks can update the same record concurrently. 3) VSAM/RLS is one way to protect the data, but that is another can of worms.

Am I understanding the z/OS side correctly?


----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to