Of course, buffering will probably do the most to improve VSAM REORG times.
For the sequential reads through the cluster that are typical of REORG
activity, add the ',AMP=('BUFNI=x,BUFND=yyy')' parameter to the DD card, where
'x' equals the number of index component LEVELS displayed in the IDCAMS
'LISTCAT ENT(<cluster>) ALL' output and 'yyy' equals the number of data
component CIs in one data CA.
This will assure that every chain of channel commands reads an entire virtual
cylinder. When the virtual read/write heads cross a cylinder boundaries, the
channel program will lose control, so no more than a cylinder can be read or
written by a channel program.
To optimize the value of 'yyy' for BUFND, look at the data component of the
'LISTC ENT(<cluster>) ALL'. Find the values for 'CI/CA' and 'TRACKS/CA'. Since
there are 15 tracks in a 3390 cylinder, you can easily calculate the number of
data CIs in one cylinder. Add one to this value and use that for BUFND.
Why add one? Some pople say the code reserves one CI for SPLIT activity, even
when the odds of it occuring are low. Some people say adding one for a CI
reserved for SPLIT activity is an old-wives tale, but it seems to work.
-----Original Message-----
From: Michael Watkins <[email protected]>
Sent: Sunday, August 30, 2020 6:00 PM
To: IBM Mainframe Discussion List <[email protected]>
Cc: Michael Watkins <[email protected]>
Subject: RE: Simple VSAM question on sizing INDEX component
The index component alone is 2.4 MB? 2,400,000/56664 = 43 tracks? Sure,
Allocate the index component as 4 or even 5 cylinders primary and a single
cylinder secondary.
If the cluster has changed significantly over time, I might also run an IDCAMS
EXAMINE command:
EXAMINE NAME(<cluster name>) INDEXTEST DATATEST
And look for:
IDC01728I FOUND n EMPTY CONTROL AREAS THAT HAVE NOT BEEN RECLAIMED IDC11775I
nnn DATA COMPONENT CIS ARE ESTIMATED TO BE UNREACHABLE
These may indicate that you want to change the index CI size, although buffer
specifications may also have to change if you do.
And I assume you have already activated 'CA reclaim' for all pertinent
DATACLASes in ISMF.
-----Original Message-----
From: IBM Mainframe Discussion List <[email protected]> On Behalf Of
Lizette Koehler
Sent: Sunday, August 30, 2020 5:12 PM
To: [email protected]
Subject: Simple VSAM question on sizing INDEX component
CAUTION: This email originated from outside of the Texas Comptroller's email
system.
DO NOT click links or open attachments unless you expect them from the sender
and know the content is safe.
List -
I have a VSAM Dataset that has grown over the years. When it was set up - the
INDEX space was left to default
I am wondering if it makes sense to override the Track Allocation and put it in
Cylinders.
We are noticing a little bit of an increase in run time during reorg. I was
wondering if this might be due to the data set having 3.4GB now. This file is
EA/EF so it can grow
Over 4500 Cylinders on the Data
And the index is using tracks 2 pri and 11 sec - size is now 2.4 MB
During reorg we offload records to an archive then reload the current data back
in.
This may be something I cannot improve on, just thought I would see if there
are any insights I am missing.
Process:
Offload the data to a temp file
Archive records older than 2 weeks
Del/Def VSAM dataset
Reload current records to VSAM dataset
This runs daily
Thank you
Lizette
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions, send email to
[email protected] with the message: INFO IBM-MAIN
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
