[OSM-talk] FAQs from help.osm.org

[Kai Krueger]

Nic Roets wrote:
> 
> I'm not sure when you last tried it. I now have a server that keeps
> the complete planet in RAM, making it considerably faster.
> 

I think it is fair to say that the KIT routing engine, which powers
routingdemo.geofabrik.de, scales considerably better with respect to route
distance on the time it takes to calculate a route. For example, I don't
think Gosmore (powering http://nroets.dev.openstreetmap.org/rails/ ) would
be able to handle a route from Lisbon to Vladivostok.

However, that isn't really the question. The question is, is it fast enough
for the available hardware for the intended purpose of validating osm data
for routing purposes. I.e. to spot missing turn restrictions, incorrect
connectivity, missing max speed tags, missing access restrictions, missing
or incorrect one way tags or other important tagging for routing that
doesn't show up on the mapnik rendered tiles. Most of those can be done
perfectly well with city length routing, which gosmore can handle just fine.

So it is imho likely that Gosmore would be sufficiently powerful on
realistic hardware for the job.

Lukily however, the patch that Frederik sent supposedly supports both the
KIT routing and gosmore simultaneously, so the user can choose the routing
engine that supports their needed features most. Furthermore, the resources
Gosmore and the KIT routing need (CPU and RAM) are possibly somewhat
complementary, so it might even be possible to run both on the same server.


Kai


--
View this message in context: http://gis.638310.n2.nabble.com/FAQs-from-help-osm-org-tp6336079p6338928.html
Sent from the General Discussion mailing list archive at Nabble.com.