At 09:26 PM 28-03-2001 GMT, Daniel ?erud wrote:
>
>I noticed it greatly affects the speed.
>Is there some way you can turn that feature off as I am
>doing speed analysis of postgresql (well, i'm kinda
>tweaking). How often should you run it? I guess it is bad to
>run it after every update.
>
>Another thought: doing _only_ inserts and selects wont
>affect the speed of the db?
>
Yeah. I also noticed that update + indexed select repeatedly on the same
row slows down over time, whereas insert + indexed select doesn't seem to
slow down for some reason. Maybe it has something to do with the indexes,
or the MVCC thingy - scanning through multiple expired rows.
The speed goes down quite significantly from the initial peak, so
benchmarks on a pristine database aren't really reliable for real world
scenarios unless you vacuum every 10000 updates or so ;). But it's a shame
because Postgres is blazingly fast at the start.
Link.