we got since a few days an performance problem
in session "Select Invoices for Reminding".

there's an select over tfacr200 and a 5 other tables
(full-tables-scan).
the tfacr200 has around 180000 rows.

we have incredible low perfomance in this session
(with standard-program). (runtime over 4 hours)

the os becomes very slow after starting this session.

i changed the script, so that the main-select is divided into
two parts, thirst selection tfacr200, after that i select
the other tables in selectdo.

after that the performance (runtime) is ok (~ 20 min).

we have an 3-tier scenario over oracle-sql-net
(ora 7.3.4 - i know that this version is desupported).

any help or ideas welcome.

thanks

rochus,

Some questions:
- can you trace what went wrong (SQL_TRACE or 10046 trace)?
- have you updated statistics?
- have you set event 100181 to 1500 or so to minimize parse time?

Probaly any of these can help.

Dick

PS what is wrong with 7.3.4?

a)
event = "10181 trace name context forever, level 1500"
in init.ora
b)
i made update statistics, but not over all,
only these tables including indexes.
c)
tt_sql i have to try again, but i can only
do that outside rush-hour.

no problem with 7.3.4, but in one of you last statements
you remebered me to that fact (don't care)

thank you

I mentioned SQL_TRACE not TT_SQL_TRACE. For detail on starting this read:
site with trace info (http://www.petefinnigan.com/ramblings/how_to_set_trace.htm) or
my tuning and trace document (http://www.baanboard.com/baanboard/showthread.php?s=&threadid=7665)

Happy tracing,
Dick