Re: URGENT HELP about 'duration' stats

Camilo Porto <camiloporto@hotmail.com> writes:
> The problem I have encountered is that the sum of executor's
> duration time is, *sometimes*, bigger than the total time interval in
> which the statements had been executed!! And this makes no sense!

Umm ... why not?  If you have, say, two queries executing in parallel
for 1 second, they'll each report a duration: of 1 second, thus summing
to 2 seconds, but the elapsed time was only 1 second.

If you don't see that always, then your benchmark program isn't trying
very hard to run more than one query in parallel ...
        regards, tom lane