if you're doing updates in a single transaction, you'll realize speed gains
by distributing the updates into multiple transactions. postgres won't have
to keep multiple copies that way.
hth
kapil
> >
> > 3) executed this statement tons of times:
> >
> > update test set data=1234 where key=1
> >
> > Here are the results -- it's pretty discouraging, I hope I'm making some
> > simple mistake, or maybe this is expected behavior for some reason?
> >
> > After this many updates ...it took this long for 1000 more updates
> > ----------------------- ------------------------------------------
> > 0 10880 ms
> > 5,000 10549 ms
> > 10,000 17380 ms
> > 15,000 20040 ms
> > 20,000 20060 ms
> > 25,000 20589 ms
> > 30,000 30749 ms
> > 35,000 30350 ms
> > 40,000 30910 ms
> > 45,000 37570 ms
> > 50,000 40379 ms
> >
> > This seems to be independent of starting and stopping my client and the
> > postmaster, running vacuum, praying, etc. I'm on RedHat6.2
> > running with the 7.1beta4 rpms.