Re: Processing Unit Time

[Charles Wilt <charles.wilt@xxxxxxxxx>]

Note that an index scan is one step down from a table scan...

It means the entire index was read.  Since the index records are
smaller than the table record, it's better than a table scan but not
by much.

You might see if VE recommends another index be created.

Charles

On Tue, Dec 21, 2010 at 10:10 AM, Schutte, Michael D
<Michael_Schutte@xxxxxxxxxxxx> wrote:
> This is from the PRTSLQINF command.
>
> SELECT SUM ( CASE : H WHEN 'DR' THEN ECDEAM ELSE ECDQTY END )
>  INTO : H
>  FROM IMSPRECH
>  JOIN IMSPRECD
>    ON ECDCONO = ECHCONO
>   AND ECDSTR = ECHSTR
>   AND ECDRCT = ECHRCT
>  WHERE ECHCONO = : H
>   AND ECHDTE BETWEEN : H AND : H
>   AND ECDCONO = : H
>   AND ECOVND = : H
>   AND ECDEXT = : H
>
>  SQL4021  Access plan last saved on 12/20/10 at 10:28:05.
>  SQL4020  Estimated query run time is 9 seconds.
>  SQL4017  Host variables implemented as reusable ODP.
>  SQL4007  Query implementation for join position 1 table 2.
>  SQL4006  All indexes considered for table 2.
>  SQL4008  Index IMSLRECD9 used for table 2.
>  SQL4011  Index scan-key row positioning used on table 2.
>  SQL4007  Query implementation for join position 2 table 1.
>  SQL4006  All indexes considered for table 1.
>  SQL4008  Index IMSLRECH1 used for table 1.
>  SQL4014  1 join column pair(s) are used for this join position.
>  SQL4015  From-column 2.ECDCONO, to-column 1.ECHCONO, join operator MF, join predicate 1.
>  SQL4015  From-column 2.ECDSTR, to-column 1.ECHSTR, join operator MF, join predicate 2.
>  SQL4015  From-column 2.ECDRCT, to-column 1.ECHRCT, join operator MF, join predicate 3.
>  SQL4011  Index scan-key row positioning used on table 1.
>
> If it was doing a table scan, the print out above would show it and wouldn't the CPU time be a lot greater.   FYI, we didn't have any other jobs running on the system with higher priority. CPU% was around 70%.  I don't know.  I'm starting to believe that it had nothing to do with the statement itself but rather the repeated calls to WrtSection of the Dev2 procedures.
>
> -----Original Message-----
> From: midrange-l-bounces@xxxxxxxxxxxx [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Morgan, Paul
> Sent: Tuesday, December 21, 2010 9:37 AM
> To: Midrange Systems Technical Discussion
> Subject: RE: Processing Unit Time
>
> Michael,
>
> Try running the process under debug (STRDBG/ENDDBG).  That will get the SQL access plan printed in the job log.  You can then review the access plan to see if it's actually using the logicals.  Sometimes SQL will reject a logical for a sequential access if the previous run was very fast.  It thinks a sequential read through the file will be faster than including the overhead of an indexed read.
>
> Paul Morgan
>
> Principal Programmer Analyst
> IS Supply Chain/Replenishment
>
> -----Original Message-----
> From: midrange-l-bounces@xxxxxxxxxxxx [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Schutte, Michael D
> Sent: Tuesday, December 21, 2010 8:37 AM
> To: Midrange Systems Technical Discussion
> Subject: Processing Unit Time
>
> I'm having trouble finding an explanation of this but I was wondering if someone could explain this here.   A user submitted a job the other day that took 6 hours to complete.  By looking at the SQL statement in the program, I saw where an improvement could be made.  I created a new logical by the selection keys in the where clause, so now the statement take less than a second to run for a month of data and the entire company.  The user resubmitted the job yesterday and this time it took 12 hours to complete while using the new logical.  If you look at the job log it says that it only used 1448 seconds processing unit time.   That's only 24 minutes.  So why did it take 12 hours.  Is it because it created a 23,000 page job log?
>
> There are two messages that are repeated over and over again that I can fix.  Dataarea CGIDEBUG not found and null indicator require.  The reason for the CGIDEBUG message is because I'm using the CGIDEV2 procedures to build an HTML page and then use write to stream file function so that I can email it to the user.  Then for the null indicator, the sql statement is summing a field but I'm not checking any statuses nor using a null indicator.  I've always just set the INTO field to zero prior to executing the statement.  I should change my habit to always use a null indicator so that this message isn't created in the job log.  But I don't see why this would cause the program to run for 12 hours.
>
> Any thoughts?
>
>
> Bob Evans makes the holidays easier. Our Farmhouse Feast is a complete holiday meal to go, ready to heat at home, serve and enjoy. Serves 6-8 for just $74.99. Pre-order through your local Bob Evans or online at www.bobevans.com/onlineordering<http://www.bobevans.com/onlineordering>
>
>
>
> --
> This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
> To post a message email: MIDRANGE-L@xxxxxxxxxxxx
> To subscribe, unsubscribe, or change list options,
> visit: http://lists.midrange.com/mailman/listinfo/midrange-l
> or email: MIDRANGE-L-request@xxxxxxxxxxxx
> Before posting, please take a moment to review the archives
> at http://archive.midrange.com/midrange-l.
>
>
> --
> This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
> To post a message email: MIDRANGE-L@xxxxxxxxxxxx
> To subscribe, unsubscribe, or change list options,
> visit: http://lists.midrange.com/mailman/listinfo/midrange-l
> or email: MIDRANGE-L-request@xxxxxxxxxxxx
> Before posting, please take a moment to review the archives
> at http://archive.midrange.com/midrange-l.
>
> --
> This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
> To post a message email: MIDRANGE-L@xxxxxxxxxxxx
> To subscribe, unsubscribe, or change list options,
> visit: http://lists.midrange.com/mailman/listinfo/midrange-l
> or email: MIDRANGE-L-request@xxxxxxxxxxxx
> Before posting, please take a moment to review the archives
> at http://archive.midrange.com/midrange-l.