Hi Laurent, I'll answer for the world wide case (for europe they should be roughly half sized):
* for CH e.g. the car profile you need 25GB for import and 18GB for running it if you use the RAM_STORE option * without CH you need 25GB to import several vehicles, running it requires only ~12GB but every request can require more RAM, see also #104 * Using bike2 or foot requires roughly 50% more for import and running, CH preparation takes also 2 to 3 times longer than with car (which is currently at 2h-2.5h) * In my tests SSD drive only slightly improves the import speed which is roughly at 1 to 1.5h * But for enabled elevation SSD is very important, also you need roughly 5GB more for import. * Using the mmap config makes everything a lot slower, I've not tested it for a recent version. Also the speed still depends of the available RAM. Using mmap for import can save you a bit RAM but using mmap for running GraphHopper is only recommended on mobile devices. * As long as you use areas which create graphhopper files not bigger than 2gb than it should be possible on a 32bit Android to still do route calculation on it. Does this answer your questions? Maybe others can throw in their experience if my values are completely off :) ! Regards, Peter On 29.01.2015 22:24, Laurent Bendel wrote: > Hi all, > just a general question. What kind of RAM does a server needs to serve > routing request on a fairly large map (say Europe) ? and what about the > complete planet file ? > I’m more interested in the amount of RAM, disk space is rather cheap, but > does it benefit from a fast drive such as an SSD ? > I guess it depends whether you use CH or not, It’d be interested to know the > difference. And if the memory mapping is used instead of RAM, what are the > minimum ? I read an Android device is capable of calculating routes, but > probably not for a very big planet file. > > Last, does the graph building process require more or less RAM than serving > requests ? > > Thanks, > Laurent > > > _______________________________________________ > GraphHopper mailing list > [email protected] > https://lists.openstreetmap.org/listinfo/graphhopper _______________________________________________ GraphHopper mailing list [email protected] https://lists.openstreetmap.org/listinfo/graphhopper
