RE: SQL vs "native" I/O performance

[Paul Conte wrote:]
The complete article explains the methodology (which was fairly rigorous)
and full results for a variety of tests. Here are a few examples: for bulk
inserts, SQL was about 12% faster than "native" RPG IV. For bulk, keyed
sequential retrieval of selected columns, SQL was about 3% faster. For
direct retrieval of a single row by key, "native" was about 55-65% faster,
depending on the number of columns retrieved. For retrieving a set of rows
with a common high order key column value (e.g., a customer's orders),
"native" was 15-20% faster.
----------

Paul, I don't want to turn this mailing list into a debate about SQL versus
native I/O, but it is very important to understand that SQL has its uses.  If
I'm wrong about SQL performance in a transaction environment, I'd love to know
why.  If I choose to set myself up as an "authority" by opening my mouth, I
should definitely research my statements.  To that end, I actually wrote a test
program and tried it before I posted my message.

Clearly, I'm absolutely astounded that your test results showed bulk inserts
faster in SQL.  Obviously, there's something wrong with my methodology.  Could
you point out how I can "fix" the following program to get results closer to
yours?  I just re-ran it several times and got results similar to yesterday: 14
seconds for SQL, 2 for native.

     FDATEFILEO   E                    DISK
     I              '1999-01-01'          C         XCDAT1
     I              '1999-12-31'          C         XCDAT2
     I              10000                 C         XCLOOP
     C                     MOVE XCDAT1    DATE
     C                     MOVE XCDAT2    DATEA
     C                     MOVE '010799'  DATA
     C*
     C                     TIME           TIMES1  60
     C*
     C           1         DO   XCLOOP
     C/EXEC SQL
     C+ INSERT INTO DATEFILE VALUES(:DATE, :DATEA, :DATA)
     C/END-EXEC
     C                     ENDDO
     C*
     C                     TIME           TIMES2  60
     C*
     C           1         DO   XCLOOP
     C                     WRITEDATEREC
     C                     ENDDO
     C*
     C                     TIME           TIMES3  60
     C*
     C                     MOVE *ON       *INLR

Thanks a million!

===================
Joe "Zappie" Pluta
www.zappie.net/java
"Where the AS/400 speaks Java with an RPG accent"
===================


+---
| This is the JAVA/400 Mailing List!
| To submit a new message, send your mail to JAVA400-L@midrange.com.
| To subscribe to this list send email to JAVA400-L-SUB@midrange.com.
| To unsubscribe from this list send email to JAVA400-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---