First time posting here.
I am attempting to run the planet-latest.osm.pbf, however am encountering a small issue. Namely it looks like idLookup time is in the range of about 6 minutes. The location_index file is about 400 MB big.
Do you have any ideas why that could be? Is there anything obvious I may be doing wrong?
I can provide more settings, but don’t know which are relevant. Code/settings-wise I am basically running the unmodified master (0.11).
How long does one routing request take?
Namely it looks like idLookup time is in the range of about 6 minutes.
This should not be. Give it enough RAM via Xmx and read also the deploy.md doc.
Thanks for getting back to me and for your suggestions. I’ll try again following everything by heart. Here’s some debug info from the server:
INFO [2018-08-06 10:11:31,597] com.graphhopper.resources.RouteResource: point=52.829321%2C12.030029&point=51.937492%2C13.535156&type=json&locale=en-US&vehicle=car 127.0.0.1 en_US Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/67.0.3396.87 Safari/537.36 OPR/54.0.2952.54 [52.829321,12.030029, 51.937492,13.535156], took:358.39682, , fastest, car, alternatives: 1, distance0: 208313.09986620065, weight0: 10204.992819555027, time0: 170min, points0: 1438, debugInfo: idLookup:358.31818s; , algoInit:0.003381019s, dijkstrabi|ch-routing:0.017237976s, extract time:0.004601574;visited nodes sum: 370;, algoInit:0.003381019s, dijkstrabi|ch-routing:0.017237976s, extract time:0.004601574;visited nodes sum: 370, simplify (2249->2249)
Additionally what I suggested above please post your config and the used GraphHopper version.
Thanks for your assistance. Checking out a latest GH code from git and using bare config you pointed me to helped! In fact my old config looked exactly like that. Anyway, the only difference I can think of (besides a bunch of updated files from git) is that I had a couple of custom profiles. Not used though. Will check that in next step.
Btw, it takes about 12 hours to get a contracted graph for a car profile for the entire planet-latest.osm.pbf - giving Java 57 GB of RAM and a machine has a total of 8 cores (AWS’s r4.2xlarge). Is this duration something to be expected nowadays?
Yes, can be ~50% faster with some dedicated Xeon CPUs but it is in this range.