Please see:
<https://www.ibm.com/docs/en/zos/3.2.0?topic=sets-characteristics-compressed-format-data>Characteristics
of compressed format data sets - IBM Documentation
The data format for a compressed format data set
consists of physical blocks whose length has no
correlation to the logical block size of the data
set in the DCB, DCBE, and the data set label. The
actual physical block size is calculated by the
system and is never returned to the user.
However, the system maintains the user's block
boundaries when the data set is created so that
the system can return the original user blocks to
the user when the data set is read.
SMF report of such files:
ID DSORG RECFM LRECL BLKSIZE SMF14CIS SMF14CMPTYPE
14 PS VBS 32760 32760 27966 2:TAILORED
14 PS VB 4096 32760 27966 2:TAILORED
14 PS FB 80 32720 56632 3:ZEDC
14 PS FB 32752 32752 27966 1:GENERIC
Michael
At 11:30 AM 6/30/2026, Michael Watkins wrote:
Please either expalin how to allocate an 'SMS
compressed file with super-blocks that write
full tracks of data' or point to some IBM documentation that explains this.
I assume from the context of your previous
remarks that 'full tracks of data' means there
are no inter-record gaps within the blocks; a BLKSIZE=56664, essentially.
I'm not assuming your previous comments don't
make sense. Despite decades of mainframe
infrastructure experience, I've never heard of
this, but I certainly don't know everything.
-----Original Message-----
From: IBM Mainframe Discussion List
<[email protected]> On Behalf Of Michael Oujesky
Sent: Tuesday, June 30, 2026 11:18 AM
To: [email protected]
Subject: Re: Net bytes per track
Nothing complex. Merely create the dataset with
SMS compression. Allocation and the SAM access
method takes care of the details.
Check you 14/15 SMF records for the actual
physical size of blocks on DASD. 14/15 still
have BLKSIZE, but is is now logical and not physical.
Michael
At 10:49 AM 6/30/2026, Michael Watkins wrote:
>Clear as mud. How do I define a data set using full tracks of
>super-blocks using zEDC compression? Of course, a DATACLAS specifying
>zEDC compression will have to be assigned, but what allocation
>parameters will use 'super-blocks'?
>
>If it's too many words to explain, please just point to a manual or red
>book that documents this.
>
>
>-----Original Message-----
>From: IBM Mainframe Discussion List
><[email protected]> On Behalf Of Michael Oujesky
>Sent: Tuesday, June 30, 2026 10:28 AM
>To: [email protected]
>Subject: Re: Net bytes per track
>
>
>Any SMS compressed data set.
>
>Michael
>
>
>At 10:13 AM 6/30/2026, Michael Watkins wrote:
> >I'll bite. '[DF]SMS compressed files super-block and write full
> >tracks of data'? What file type, accessible through any standard z/OS
> >access method, can use 'full track super-blocks'? Please explain.
> >
> >-----Original Message-----
> >From: IBM Mainframe Discussion List
> ><[email protected]> On Behalf Of Michael Oujesky
> >Sent: Tuesday, June 30, 2026 9:46 AM
> >To: [email protected]
> >Subject: Re: Net bytes per track
> >
> >
> >Not quite correct. SMS compressed files super-block and write full
> >tracks of data. BLKSIZE becomes logical (i.e. the buffer size of
> >data used between the applcation and the SAM access method.
> >
> >See SMF14CIS in the 14/15 SMF record for the actual physical size on
> >the track.
> >
> >Michael
> >
> >At 08:19 PM 6/29/2026, Michael Watkins wrote:
> > >Content-Transfer-Encoding: base64Yes, 56664 bytes per track on a
> > >3390, but the maximum z/OS BLKSIZE is 32,760 bytes and only one
> > >block will fit on a track.
> > >
> > >To maximize the amount of data on a track, half-track blocking is
> > >typically used. There is a 668-byte inter-record gap on a native
> > >3390 (platters on a spindle), which has been replicated on RAIDed
> > >FBA devices in modern DASD architecture. This means there is a
> > >maximum half-track BLKSIZE of 27998 bytes since 668+(2x27998)=56664.
> > >
> > >For a data set with 80-byte records, this means half-track blocking
> > >means a 27920-byte BLKSIZE since 349 x 80 = 27920 and there isn't
> > >room in
> > >27998 for the 350th 80-byte record.
> > >
> > >Also keep in mind that z/OS will append 32 bytes onto each block
> > >when DSNTYPE=EXTENDED is specified, meaning less space for data
> > >records is available.
> > >
> > >
> > >
> > >-----Original Message-----
> > >From: IBM Mainframe Discussion List <[email protected]> On
> > >Behalf Of Joseph Reichman
> > >Sent: Monday, June 29, 2026 8:01 PM
> > >To: [email protected]
> > >Subject: Re: Net bytes per track
> > >
> > >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.
> > >
> > >I keep on seeing from AI 56,664 bytes per tack for 3390 pack how is
> > >that determined
> > >
> > >
> > >Joe Reichman
> > >
> > >On Mon, Jun 29, 2026 at 8:58âEUR¯PM Attila Fogarasi <
> > >[email protected]> wrote:
> > >
> > > > The net bytes (i.e. longest record you can write to that track)
> > > > varies by access method as there is overhead (control bytes).
> > > > TRKCALC is provided to compute the largest record which can fit
> > > > on a track (using EXCP), see
> > >https://urldefense.com/v3/__http://www.ibm.com/docs/en/zos/3.2.0?to
> > > >
> > > > Unless you are using EXCP the hardware track capacity is
> > > > irrelevant,
> > >HXØÙ\ÜÈY]Ù\ÙYÚ[]@rmine the actual bytes per track (and
> > > > varies by BLKSIZE).
> > > >
> > > > On Tue, Jun 30, 2026 at 10:35âEUR¯AM Joseph Reichman <
> > > > [email protected]> wrote:
> > > >
> > > > > Would anyone know how to get the net bytes per track for a
> > > > > 3390
> > > > >
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
