[prev in list] [next in list] [prev in thread] [next in thread] 

List:       sas-l
Subject:    Re: JCL SPACE ALLOCATION ISSUES
From:       pudding man <pudding_man () MAIL ! COM>
Date:       2004-02-28 17:08:28
[Download RAW message or body]

As nearly as I can tell, STR has 55m recs of length 242
to process. 55 x 242 = 13+ gb. Cranking 13+ gb of data
around on the average 'mainframe' system is not trivial,
even for an experienced programmer. Efficiency consider-
ations can become critical.

Suggestions are as follows:

1.) Repost describing the data and what you need to do
    with it in detail. Include SAS code and any relevant
    SAS log segments. There are many folks here who can
    help.
2.) Begin program development using a small subset of the
    data. Perhaps something like OPTIONS OBS=50000. You
    can always delete/re-allocate necessary datasets with
    more space. Keep in mind the final data volume
    requirements.

Avoid running on huge datasets unless you have solid
expectations that the program will run correctly (else
you'll just waste resources).

Count bytes, and get familiar with relevant devices
(probably 3390 DASD) at your 'mainframe' site. You'd
likely get good mileage from something like "... JCL"
by Gary DeWard Brown.

After you have developed/tested good, efficient program(s),
you can then engage the "hounds" (DASD Police) as necessary.

Hope it hep's ...

  Skoal,
  Puddin'

*******************************************************
***** Puddin' Man **** Pudding_Man-at-mail.com ********
*******************************************************;

"It is often noted that our monetary systems make barter largely
 obsolete. While this is true for small purchases, very large
 purchases (i.e. a Senators vote, a Presidents policy) are often
 bartered through spheres of influence. In such transactions,
 cash may be used but is viewed by most as a 'no-no' due to
 certain legal considerations. Some believe that human lives
 are more-or-less bought/sold on a daily basis by means of such
 barter."


STR wrote:

> I am running into what appears to be space allocation issues
when I
> attempt the following:
>
> I am trying to read in a massive file, and break it up into 3 or
4
> separate files based on
> the record type for the purposes of later examination.  I am not
a JCL
> guy, but I researched
> the DD statement parameters to gain an understanding of a way to
best
> achieve this.
>
> I am using mainframe SAS to split the files.  I know that this is
> coded correctly.  The
> errors I receive appear to relate to the JCL, specifically space
> allocation issues.  The
> maximum record length of the input file is approximately 242.  I
am
> projecting approximately
> 55 million records that qualify for one of the record types.
>
> I should note that from the input file I only require maybe 10
fields
> or so.  Currently, I
> am placing that target data in the same positions as reflected
in the
> input file to preserve
> the original record layout.  However, I am thinking this is
probably
> wasteful of space since
> a number of the fields I am not concerned with and contain
spaces.
> So, I believe this is one
> area where I can cut back on required space allocation.  So is
this
> done by simply recoding
> the LECL parameter to the new record length required for my
target
> file based upon only the
> fields I am interested in retaining?  And then I guess my SAS put
> statement would need to be
> revised to specify positions that scrunch the data closer
together
> rather than placing the data
> in the same field position as the original file?
>
> Resolution to that might possibly resolve my problem.  I am just
not
> sure if that is the way
> to go.
>
> As mentioned above, the first dataset I would like to create will
> contain approximately
> 55 million records.  Here are some of the DD statements that I
have
> used so far which have
> brought errors:
>
> //OUTPUT1  DD DSN=DATA.SET.NAME,DISP=(,CATLG,DELETE),
>               DCB=(LRECL=242,RECFM-FB),
>             UNIT=DISK,SPACE=(CYL,(300,30),RLSE)
>
> The error that resulted:
>
> ERROR:  OUT OF SPACE WRITING TO THE FILE OUTPUT1.
> FATAL:  UNRECOVERABLE I/O DETECTED IN THE EXECUTION OF THE DATA
STEP
> PROGRAM. ABORTED DURING
>         EXECUTION PHASE.
>
> Only a portion of the records were saved out.
>
>
> I also tried the AVGREC approach:
>
> //OUTPUT1  DD DSN=DATA.SET.NAME,DISP=(,CATLG,DELETE),
>               DCB=(LRECL=242,RECFM-FB),
>             UNIT=SYSDA,
>             SPACE=(242,(55,2),RLSE),
>               AVGREC=M
>
> The error that resulted:
>
> ALLOCATION FAILED DUE TO DATA FACILITY SYSTEM ERROR
> ALLOCATION FAILED FOR DATA SET DATA.SET.NAME, PRIMARY SPACE
EXCEEDS
> 65535 TRACKS
>
> Step one was not executed.
>
>
> I'm just leary about being too over-zealous in allocating space.
 I am
> reluctant in just
> going with trial and error since I am not a mainframe guy and am
not
> quite sure if I could
> cause some performance issues or if the space I am asking for
will
> send the hounds after me.
>
> Also, I'm not sure to the extent that the working files matter,
but I
> have allocated about 5
> working files each defined as follows:
>
> //WORK     DD   UNIT=SYSDA,SPACE=(CYL,(500,500),RLSE)
> //WORK1    DD   "..................................."
>
> Any advice/direction would be greatly appreciated.

--
___________________________________________________________
Sign-up for Ads Free at Mail.com
http://promo.mail.com/adsfreejump.htm

[prev in list] [next in list] [prev in thread] [next in thread]