reduce running time of a jcl

hyuzen · Active User Joined: 10 Jul 2005 Posts: 50

Dear Frank,
Actually I need your advice.
Here is a quote of a jcl which processes over than 50 mio records.

//STEP1 EXEC PGM=IDCAMS
//SYSPRINT DD SYSOUT=*
//SYSIN DD *
DELETE PTRC.ITEMIZED.ESDS
DELETE PTRC.ITEMIZED.KSDS
DEFINE CLUSTER( -
NAME(PTRC.ITEMIZED.KSDS) -
INDEXED KEYS(32 0) -
RECORDSIZE(98 98) CISZ(16384) -
CYLINDERS(1000 100) -
) -
DATA( -
NAME(PTRC.ITEMIZED.KSDS.DATA) -
) -
INDEX( -
NAME(PTRC.ITEMIZED.KSDS.INDX) -
)
/*
//STEP2 EXEC PGM=IEFBR14,COND=EVEN
//ALLOC1 DD DSN=PTRC.ITEMIZED.ESDS,DISP=(,CATLG),
// RECORG=ES,LRECL=98,BLKSIZE=0,
// UNIT=3390,SPACE=(CYL,(1000,100))
/*
//STEP7 EXEC PGM=ICEMAN,PARM='SIZE=MAX',REGION=7M
//SYSPRINT DD SYSOUT=*
//SYSOUT DD SYSOUT=*
//SORTIN DD DSN=PTRC.ITEMIZED.ESDS,DISP=SHR,BUFNO=200
//SORTOUT DD DSN=PTRC.ITEMIZED.KSDS,DISP=SHR
//SYSIN DD *
SORT FIELDS=(1,32,CH,A)
RECORD TYPE=F,LENGTH=98
OPTION FILSZ=E10000000,DYNALLOC=(WORK,18)
/*

I want to reduce running time of step7. What do you suggest ?

Thanks in advance.
hyuzen

Frank Yaeger · Posted: Wed Sep 14, 2005 3:20 am Post subject: Re: reduce running time of a jcl

I can't really comment on your question without seeing the DFSORT diagnostic messages for a successful run which requires adding the following to your job and rerunning it:

//SORTDIAG DD DUMMY

Assuming you're using DFSORT, if you want to send me a note offline (yaeger@us.ibm.com) early tomorrow morning with your JCL, control statements, JES messages and //SYSOUT messages, I'll forward it to a colleague who's actually better at performance and tuning than I am so she can take a look (I actually won't have any internet access after mid-morning tomorrow for about 10 days.) If you do e-mail me, please put "DFSORT" somewhere in your Subject line to catch my attention.