A couple months back I wrote a benchmark script which runs a variety of
operations and records the results. I have come up with some interesting
results (like plain PLSQL cursor loops being faster than bulk update!) and
it is not very complicated. I guess I am wondering why something running
multiple statements like...
create big table, full scan big table, update big table, truncate big table
etc...is not enough to stress your IO system. This would be very easy to
write and isn't very complicated at all.
Obviously this doesn't simulate the load of an OLTP environment, but it
should be sufficient to fully stress IO channels.
You should record redo info, archive info and file stat info to get a
measure of performance. Statspack is fine also for some of this and getting
wait info. Play around with your IO related parameters and run a lot of
tests.
Maybe Orion does this, can someone post a brief review (Henry?).
--
http://www.freelists.org/webpage/oracle-l
Received on Fri Nov 18 2005 - 11:34:27 CST
