[OSM-dev] Osmosis Replication Statistics

[Brett Henderson]

I've just uploaded osmosis version 0.11 
(http://www.bretth.com/osmosis/osmosis-latest.zip).

It makes some changes to the database snapshot tasks to simplify the 
queries.  I'm unsure if it will make the dump quicker because it results 
in much more data being returned to the client and larger temporary 
files being produced.

I've also included gzip and bzip2 support to directly read and write 
compressed files.  Gzip seems pretty fast using the standard JDK which 
uses some native methods internally, bzip2 may be considerably slower.  
If raw speed is an issue it will almost definitely be faster to use 
separate command line tools.

Next up I may try to generate a delete changeset for the problematic 
original tiger data.  I don't know how this should applied to the 
database but it is the same problem the new TIGER import and AND data 
are currently facing.

Brett Henderson wrote:
> I just produced 20070721.osm by applying 20070720-20070721.osc to 
> 20070720.osm.  It took 43m34s to complete.
>
> It took 3m07s to produce 20070720-20070721.osc so this means that a 
> daily planet could be produced in approximately 1 hour with less than 5 
> minutes of database activity.  The reality on a production system might 
> be slightly different of course.
>
> I'll take another look at the snapshot task.  Based on IRC discussions 
> there may be a better way of doing this.
>
>
>
> _______________________________________________
> dev mailing list
> dev at openstreetmap.org
> http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/dev
>