On 06/21/2012 07:47 PM, Joel C. Ewing wrote:
On 06/21/2012 06:42 PM, Lizette Koehler wrote:
From: "Joel C. Ewing" <[email protected]>
And better yet, also use a DATACLAS with the Extended attribute so you
are not constrained to only 16 extents per volume.
With Space Constraint Relief. Extended attribute, and Dynamic Volume
Count, both primary and secondary allocations will use as many extents
and volumes as necessary to get the requested primary or secondary space
(not just the 5 extents for primary, 1 for secondary limits of ordinary
sequential dataset allocation), so fragmentation of free space is less
of an issue. Since you are not guaranteed any minimal amount of space
for any individual extent, you also need the larger extent count limit
given by the Extended data set attribute to insure that you are not
artificially restricted by number of extents from allocating available
free space on the assigned volumes.
Using these attributes you can even request a primary allocation that
spans multiple volumes and guarantee that step execution will not even
start if a large minimal primary space is unavailable, but you have to
be careful using this because at least in the past the RELEASE option
would not release allocated but unused space on volumes that contained
no written data, only on the last used volume of the data set (so a
multi-volume primary allocation used for a very small dataset could
consume a large amount of pool space on subsequent unused volumes even
after RELEASE).
Joel C Ewing
Joel, I am not sure that would apply to a PDS. I know that for
nonvsam sequential files that will work, but I am not sure if it is a
PDS if it can go beyound 16 extents. I have been looking through the
manauls and cannot find a reference (info center - yuck).
Can you comment?
Lizette
No, I'm pretty sure my remarks only apply to Extended Sequential
datasets and that there is no "Extended" support for PDS or PDS/E data
sets. VSAM datasets can also be Extended, but they have their own
allocation rules.
I interpreted the context of the original question to be about
sequential datasets since multiple GDG datasets were being merged -- and
the ability to create and work with GDG datasets that happen to be a PDS
is not really documented and not generally known outside those of us
that needed the capability, tried it, and found it worked (obviously you
can't specify both a member name and a relative GDG number in JCL
because of syntax restrictions). I gathered from discussion at one of
the SHARE Q&A's over a decade ago that PDS/E's can't be a GDS because
the PDS/E designers didn't appear to be aware that was one of the
"features" of a PDS and the PDS/E implementation explicitly disallowed
such usage.
And I forgot to mention earlier that a simple way to test whether your
sequential datasets are being allocated as expected is to set up a job
to IDCAMS REPRO a moderately sized sequential dataset concatenated to
enough copies of itself to build up a sequential dataset equivalent to
about half a drive in the pool, and then use that concatenated to itself
enough times as input to another REPRO to convince yourself that extents
for a multi-volume output dataset are being allocated as expected and
multiple volumes are being used -- being careful to delete all your
"junk" datasets after inspecting the results. There is an IBM utility
designed explicitly to create a specified number of test records for a
sequential dataset, but my recollection is that it is incredibly slow
and a CPU hog - much faster to use IDCAMS REPRO.
--
Joel C. Ewing, Bentonville, AR [email protected]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
