[osmosis-dev] Osmosis use case
gwalschlager at rtlogic.com
Fri Mar 26 12:18:02 GMT 2010
Again...thanks for the information!
The disk space requirements of 150GB max would be reasonable for my purposes. You mentioned that osm2pgsql needs more RAM (16GB+ during import). Do you know how much RAM would be required for normal lookup and rendering usage with either slim or non-slim usage? Is the RAM constraint only an import constraint? Bottom line is that my application will need to run on an average dual core PC with likely 4GB of RAM, so I am looking for a local database solution that will allow me to get decent performance on this kind of machine. Is this possible?
Also, I think a schema that didn't have history would be better for my purposes (less data) as I only need lookup and rendering capabilities (i.e. read-only).
From: Frederik Ramm [mailto:frederik at remote.org]
Sent: Friday, March 26, 2010 6:09 AM
To: Walschlager, Gerard
Cc: osmosis-dev at openstreetmap.org
Subject: Re: [osmosis-dev] Osmosis use case
Walschlager, Gerard wrote:
> I appreciate your quick response! I am not surprised by your answer.
> Any idea on how much disk space would be required by a PostGIS
> database that would hold the entire planet.osm file data?
I think it would be between 100 and 150 GB for either schema; but only
30 GB if you do option 2 (osm2pgsql) in non-slim mode. In non-slim mode,
osm2pgsql needs more RAM (at least 16 GB during import) but produces a
smaller, and non-updatable, database.
> I'm not sure what the "simple schema" is. I assume that is different
> than resolution #2 that you indicated in your reply?
Yes, osmosis and osm2pgsql use different schemas. Osmosis supports the
APIDB schema and the "Simple" schema while osm2pgsql has its own schema,
available in "slim" and "non-slim" varieties.
From a data modeling perspective - and I'm on somewhat thin ice here,
happy for anyone to chip in and confirm or protest - the APIDB schema is
the most powerful because it has all the data plus history. The "Simple"
schema, as well as the Osm2pgsql slim-mode schema, still have a full OSM
data snapshot (and you could theoretically convert from one to the
other, or from APIDB into one of them) but they don't have history. The
non-slim osm2pgsql schema, lastly, carries only a processed selection of
OSM data (exactly what is needed for the map - nothing more). You could
theoretically convert from any of the others to non-slim osm2pgsql, but
never in the other direction.
> I assume that
> in answer #2, there is a standard schema that Mapnik assumes must be
> in place in the database it will query and that schema is different
> than the "simple schema" specified in answer #1?
Yes. To me more precise, Mapnik itself can be made to work with any
schema, even the "Osmosis simple" schema, but the schema created by
osm2pgsql and the style sheet commonly used in OSM work together and if
you were to use Mapnik with the Osmosis simple schema you'd have to
write your own style sheet.
More information about the osmosis-dev