The purpose is to provide an environment for computing the distances between municipalities from France. For that, I propose to you to build a docker image that will install all required software and tools.
The first step is to built the docker image that will install the OSRM backend, the python library osrm-c as frontend and the scripts for the distances matrix computation.
docker build -t distances .
Now, you can start the docker image with this command:
docker run --interactive=true --tty=true --ulimit memlock=68719476736 --privileged --volume=$(pwd):/root/distancesmatrix/data distances
We need to enable the shared memory.
According to this document, we use the option --ulimit of docker with attributes memlock=68719476736
Then, as you are in the container, you have to run the script run.sh with your osm.pbf extract file as input:
./data/run.sh data/cantal-rectangle.osm.pbf
After that, the data have been extracted by OSRM and loaded into shared memory.
python data/computeDistances.py
In France (metropolitain), we have 36571 municipalities, giving 668700735 distances to compute for each couple of municipalities for one direction (municipality A to B) and 1337401470 (1.34E+9) if we consider the two directions (distance from A to B and from B to A). Considering the two directions can be determinant for some municipality, if for example they can enter easily on a motorway in a direction, but should made a detour in the other one.
Considering only couples of municipalities that have an euclidean distance lesser than 100km, we have 37101055 distances to compute (for each direction).