I am running a data warehouse application on Oracle 8.0.3 (on an 8-cpu Sun
Ultra 4000
process from flat files to
warehouse tables (via PL/SQL procedures), the
though log switches occur every 4 minutes, there are about
checkpoints between log switches. 6 different load streams are running
simultaneously.
being inserted into the transient
and they are truncated prior to the load). The
425,000 rows (200 Mb) being inserted into the data warehouse
entire load process is taking 40 minutes. MPSTAT shows that 30-40% of time
is spent in usr/sys mode, 50-60% time is spent in wait mode, and almost no
idle time.
files (100 Mb each)
log_buffer = 41943040
log_checkpoint_timeout = 0
log_simultaneous_copies = 16
1) Why are my log switches occuring so frequently (given the size of my redo
log files)?
log switches?
result in redo log entries?
specified result in 200 Mb of redo log
flushed to disk? When it reaches 1/3 full?
triggered? Is the log_checkpoint_interval parameter
db_block_size parameter to determine when this occurs?
change to reduce the number of interval checkpoints?
recommendations to improve overall load time?
-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/ Now offering spam-free web-based newsreading
Received on Wed May 20 1998 - 09:02:18 CDT
