Hi,

I'm trying to implement waterway routing over the world.

I used osmium tags-filter --progress -v -o data-filtered.osm.pbf planet.osm.pbf 'nw/waterway' --overwrite
The output file data-filtered.osm.pbf is 7Gb.

I tried to run the command osrm-extract with a custom .lua profil on EC2 with the following configuration:

• t3a.2xlarge (32 Gb RAM) : Memory overflow
• f1.2xlarge (122Gb RAM) : Memory overflow
• r4.8xlarge (244 Gb RAM ) : Working fine

I am a bit surprised. I thought the thumb rule to run the osrm-extract command was to have around file_size*5=7*5=35 GbRAM in order to run the command. I used (as you can the on the image bellow of the EC2 RAM percentage utilization) up to 50% of the 244 Gb of RAM available, so around 122 Gb to run osrm-extract for A 7Gb file.

• Is it a normal behaviour?
• Does osmium tags-filter can help reducing osrm-extract time ?