OpenStreetMap
Took my first osm xml file download yesterday – rhone-alpes.osm from CloudMade, 52m lines in 3.95Gb (uncompressed) so hardly planet.osm. Even so the parsing program needs large file support “cc -D_FILE_OFFSET_BITS=64”
First steps taken to parse out the node lat&lon pairs into array of doubles. Highest node id in rhone-alpes is 1.543.458.302 = 1.54Gnodes. Assuming this is not outrageously different from the number of nodes in planet.osm, we could store the lat&lon pairs for the planet in ~25Gb (2x8 bytes x 1.5Gnodes). That’s definitely do-able.
17m nodes of rhone-alpes loaded in 62 seconds on bog-standard laptop, so ~1.5hrs for the planet?
Discussion
Reäksje fan giggls op 26 maart 2012 om 10.19 oere
For new software pbf format will always certainly be the better choice than osm format.
Reäksje fan Jochen Topf op 31 maart 2012 om 12.18 oere
You can store the coordinates as integers and you’ll only need half the space. See the storage/byid stuff in Osmium on how this can be done. The reason this works is that the precision of the coordinates is limited anyway, because internally the OSM server also stores the coordinates as integers.