At 23:15 +0800 on 08/26/2006, Johnny Luo wrote about Re: Z/os
Performance issue: REWRITE a sequential data set:
Because the pl/i program is simple, i'll put its main logic here:
(It's our customer's production program)
DO WHILE (!EOF_FILEA & !EOF_FILEB);
DO WHILE (!EOF_FILEB & FILEA_ACT<=FILEB_ACT);
IF (FILEA_ACT = FILEB_ACT) THEN DO;
FILEB_AMT = FILEA_AMT;
REWRITE FILE(FILEB) FROM(FILEB_REC);
END;
READ FILE(FILEB) INTO(FILEB_REC);
END;
READ FILE(FILEA) INTO(FILEA_REC);
END;
From the above, the "Record Length Change" hypothesis is shoot down.
All you are doing is rewriting all the records in FileB that are in
FileA (with FileA's AMT field replacing the contents of the FileB
record's AMT field). Are you using BSAM or QSAM as your Access
Method? It should be QSAM which only flushes updated blocks after you
are done with the block. In BSAM the block gets flushed at update
even though more records remain it in (thus your high IO count since
you are updating ALL FileB records that exist in FileA.
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
