[postgis-users] Benshmark postgis

[strk]

On Fri, Nov 26, 2010 at 09:15:27AM +0100, Arnaud Vandecasteele wrote:
> Thanks you for your answers.
> I can understand your objections but how can I make this process better ?
> 
> Also, my tests gave me some strange results.
> For the first set, I did the process explain before without a gis index (gist).
> And for the seconde one I built a gis index.
> I was expected better result for the second set, but it hasn't been
> the case. Even sometimes it took me more times than without an index.
> 
> Do you know what could explain this behaviour ?

Show the numbers!
When you have an index the planner has to take a decision as to
whether or not to use the index. This decision may explain an overhead
you're seeing. If it chooses to NOT use the index the additional time
spent in deciding will be clear. If it chooses to use the index then
the speed difference can be due to other things.
First, having just a few geoms (world countries) doesn't really show
much advantage of the index.
Second, having countries spannign the whole world (have bounding boxes
shown on a map to see) doesn't make the index very efective (still finds
many candidates).

Your procedure is fine to check how good postgis is vs. others 
_given_the_exactly_same_input_
I suggest making the points predictable too rather than rendom, if
you're aiming at this kind of comparison.

If you want to evaluate something else make it clear what the goal is.

--strk;

  ()   Free GIS & Flash consultant/developer
  /\   http://strk.keybit.net/services.html