[osmosis-dev] osmosis performance optimization

[Aurélien FILEZ]

Hello,

I have an ubuntu server installed on a computer which have an i7 processor
(4 physical CPU, 8 logical) with 8Gb of memory.

My objective is to extract a country data with completeWays &
completeRelations = true, to have a better cutting than what we can have
with the geofabrik cutting. For the moment, I can make it, but it's long,
very very long (9/10 days of processing for the bounding-polygon
operation). I'm sure I can improve this treatment by a better osmosis
usage, but I don't understand stories about pipes, buffer etc.

Here is how I actually process (for example, for the France country) :
- Download european set (10.2 Go)
- Apply a "bounding-bbox" of the country (2.7 Go)
- Apply a "bounding-polygon" of the country (1.5 Go)

The commands provides from a basic usage of osmosis :
--read-pbf file=xxx.pbf --bounding + options --write-pbf yyy.pbf

I first remark that with htop, I have 8 processors, but only one used.
Maybe we can exploit the all computer performance by a "spacial command" ?

Also, I don't need all osm data, and I know exactly which entity / tags I
need, for example : Ways (tunnel, bridge, oneway, ref, z_order), Points
(addr:housenumber, addr:housename, addr:street, addr:postcode, addr:city),
Relations (admin_level), etc.

Can you, please, help me to find the better steps / commands to make this
job in the spirit of "what I need, how I need, where I need, when I need,
not too much, not too less" ?

Thank you for your help, very much, and sorry for my bad english (I will
improve it ;)

See you later,

Kin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/osmosis-dev/attachments/20130601/52f2bac6/attachment.html>