[OSM-talk] Server slowness

[Nick Black]

Time gentlemen,

Martin, you came to the list suggesting that we take a look at
PostGIS, pointing out its spatial indexing features and that its
strengths lie in storage and serving of data and that we would be
pleased with the results.  As it turns out, many people on the list
work with PostGIS professionally and some have experimented with using
Geometry types to store OSM data.  The results of peoples tests have
suggested that using Geometric types would actually slow OSM down.

The discussion then switched to the other benifits of PG,
interoperability, rubustness, rendering from Mapnik etc etc - there
are a lot of benifits.  People have explained the benifits of keeping
the existing structure, and Steve has said on the list that he'll
switch the database tomorrow - if anyone can show that there would be
benifit in doing so.  Surely this isnt unreasonable?  Imagine the
flames on the list if we switched to PostGIS so that we could be
interopable with OGC, render quickly, ship data in Shapefiles, TAB
files, import TIGER etc etc, but querying a node takes three times
longer.  People would be pissed to put it mildly.  There has to be
good reasons for switching.  You have provided a lot of good reasons -
but not the key to the problem - speed.  I am genuinely interested to
see what the restults are - I like working with Postgres and indulge
in the ease of tools like ogr2ogr - but OSM is about people being able
to put their data in and get it out as quickly as possible.  The ease
of use of the OSM API as well as the greate community here and genuine
openness, is what has spawned the numerous editors, conversion utils
and like that have made OSM take off.  Seriuosly, check out utils/
from osm svn - its a goldmine of scripts that are letting people do
all kinds of stuff with OSM data - including conversion to shp, GML
and PostgreSQL.

So again, I'm really intereseted to see how well PostGIS performs on
OSM queries.

Nick
(owner of a refractions / PostGIS tshirt)



On 1/16/07, Martin Spott <Martin.Spott at mgras.net> wrote:
> SteveC wrote:
> > * @ 16/01/07 06:42:31 AM Martin.Spott at mgras.net wrote:
> > > Please tell me when you're ready and
> > > we'll try to discuss this in a more reasonable way.
> >
> > The barrier to entry is to just prove your point. It's that simple.
> >
> > Please come back with timings showing your schema is faster than mysql
> > as it stands.
>
> I take this response as a proof that you are probably unwilling or
> simply incapable of understanding the point why I decided to
> participate into this discussion. If you've simply lost the track then
> I'll write a summary for you, if you like (but don't expect me to do
> this immediately as I have a day job to work for).
>
> Still I'm open - as promised - to benchmark different storage
> architectures,
>
> Martin.
> --
>  Unix _IS_ user friendly - it's just selective about who its friends are !
> --------------------------------------------------------------------------
>
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
>


-- 
Nick Black
--------------------------------
http://www.blacksworld.net