RE: Using PQ in FTS

Amir

> How much "CPU time" was spent by a query that was run
> in parallel with "x" number of slaves?

You have to sum the CPU time of the QC and the CPU time of all slaves. Since it is 9.2, and therefore the statement executed by the QC is different from the statement executed by the slaves, you have to do it "manually". In 10 it would be done by tkprof...

> What was the actual "elapsed time" of a query that was
> run in parallel with "x" number of slaves?

Take the elapsed time of the QC.

> If the "disk" statistic is the aggregated statistic
> obtained from the QC and all slave processes then why
> is it much smaller than that obtained from the
> serialized statement?

If you sum the disk statistic of the QC and of all slaves, it should be the right one. As far as I remember I saw differences only in two situations:
- when the execution plan was different (except for the PX part, of course...)
- because of usage of temporary tablespace (the PGA allocated to a single process is likely smaller than the sum of all PGAs allocated to the slaves, i.e. in some situation only the serial execution spills data to temp while)

HTH
Chris

--
http://www.freelists.org/webpage/oracle-l