To execute REXX faster

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

I've created a rexx routine to format an input file and the input file have 300,000 records so make not able to excute all at once
if so it gives me error message as "Machine storage exceeded"

So I split it into 50,000 records each and trying to format it, but this takes me lot of time to save more than half an hour and even more to complete
is there any way that I can make it faster.

Craq Giegerich · Posted: Thu Jul 24, 2008 5:17 pm

Use something other than REXX.

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

but i have created the REXX for this will not be able to change now.

so any other possibility in REXX?

expat · Posted: Thu Jul 24, 2008 5:32 pm

What do you mean reformat the input file ?
Just selecting various fields from the file ? Please give more details.

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

i will have a copybook
A PIC x(10)
B PIC X(5)
my inout file will be
123%rf%
1234%refgg%
so i need to format this input file for the length specified in the copybook
and the output will be like

123 rf
1234 refgg

i did the rexx for this and its working fine my concerns is if i give more than 50000 records and try executing it online its taking more than a hour so is there any way to execute this fast

PeD · Posted: Thu Jul 24, 2008 5:52 pm

Rexx is not very good for I/O,- I mean is very bad - but in any case execute that program in batch mode, execept you want to drink coffees and coffees in front of your SYSTEM screen.

expat · Posted: Thu Jul 24, 2008 6:06 pm

PeD · Posted: Thu Jul 24, 2008 6:19 pm

No Martin.

Coffee machine is out of order !! and no more biscuits !!!!

Where are you?

expat · Posted: Thu Jul 24, 2008 6:31 pm

PeD, have messaged you offline.

Robert Sample · Posted: Thu Jul 24, 2008 6:33 pm

vidyaa, I think the consensus is that if you're concerned about performance, REXX is not the way to go. If you're forced to stick with REXX, be prepared to wait for the processing to complete.

dbzTHEdinosauer · Posted: Fri Jul 25, 2008 4:24 am

if rexx is your only media,
swap the execio logic for allocate, get, put, (or whatever the read is) etc.....

dbzTHEdinosauer · Posted: Fri Jul 25, 2008 4:30 am

you want a process that is not restricted to file size.

So, read them one at a time,
the execio when you start reading mannnny records
(like this silliness that you have)
ends up spending a lot of time finding memory to hog from other services.
With the allocate you can control how much 'buffer' is being used.

superk · Posted: Fri Jul 25, 2008 8:25 am

I'm curious, does anyone know? Would the file I/O times be better if you were to use the ISPF LMGET/LMPUT services instead of EXECIO?

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

Can any one please tell me how to execute this in the batch mode. i have 2 macro involved in the REXX

am using the below JCL but that gives me some abend like S322

HappySrinu · Active User Joined: 22 Jan 2008 Posts: 194 Location: India

might help..

Add, Region, Time parms in your job card

just from quick reference...

322

Explanation: One of the following occurred:

o The system took a longer time to run a job, job step, or procedure
than the time specified in one of the following:

- The TIME parameter of the EXEC or JOB statement

- The standard time limit specified in the job entry subsystem

o For a started task under the master subsystem, the TIME parameter was
not specified on the PROC statement of the catalogued procedure, and
the PPT entry did not indicate a system task

expat · Posted: Fri Jul 25, 2008 12:19 pm

I would speak to your peers and see which job classes allow more CPU resource.

And if possible - perform some check to verify that the step is not looping before attempting to add TIME parameters in the JCL.

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

yes ..let me try that

Garry Carroll · Posted: Fri Jul 25, 2008 1:14 pm

Just spotted this thread....

Curious - the files looks to me like a .CSV type file, delimited by percent signs?

DFSORT using PARSE and JFY might be the way to go....

www-304.ibm.com/systems/support/storage/software/sort/mvs/tricks/pdf/sorttrck.pdf

Garry.

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

yes the input is delimted by % . am parsing those variables read the copybook to find the length of the varibale and if the length is same as specified in the copybook then writing it as it is else pad the difference in the length of the field and the copybook and fill it the remaining length with spaces and write to the output file.

my concerns is this runs fine online when i try with 50 to 100 records but am not able to run the same for 50,000 recordds and more. i tried with both online and batch it is taking hours and hours so wanted to make it fast

Garry Carroll · Posted: Fri Jul 25, 2008 2:37 pm

Try the following DFSORT batch job, I believe it will do what you want much more efficiently than REXX, and it's probably easier to maintain.

I assumed padding on the right for both fields (JUSTIFY=LEFT). You could alternatively specify JUSTIFY=RIGHT

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

am getting S000 U0016 on running this please advice also i have 3lakhs records like 123|~rf|~ etc.(actually its a |~ seperator not %) then how do i run this and get this formated into an output file

expat · Posted: Fri Jul 25, 2008 3:18 pm

Post the output showing the error

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

I ran the same jcl what is posted above The erro messaage
05.51.53 J0008571 NE1034RN ENDED AT HTF1 - ABENDED S000 U0016 CN(INTER
AL)
WER268A INREC STATEMENT : SYNTAX ERROR
WER211B SYNCSMF CALLED BY SYNCSORT; RC=0000

expat · Posted: Fri Jul 25, 2008 3:25 pm

Aaaaaaaaah, Garry give you syntax for DFSORT, you execute SYNCSORT

vidyaa · New User Joined: 02 May 2008 Posts: 77 Location: chennai

//STEP01 EXEC PGM=ICEMAN is what i used .
am not getting what is to be done on this regard