Running OSRM

Daniel J. H edited this page Dec 15, 2017 · 83 revisions

How to run the tool chain

There are two pre-processing pipelines available:

  • Contraction Hierarchies (CH) which best fits use-cases where query performance is key, especially for large distance matrices
  • Multi-Level Dijkstra (MLD) which best fits use-cases where query performance still needs to be very good; and live-updates to the data need to be made e.g. for regular Traffic updates

Our efforts are shifting towards the MLD pipeline with some features seeing MLD-specific implementations (such as multiple alternative routes). We recommend using the MLD pipeline except for special use-cases such as very large distance matrices where the CH pipeline is still faster.


For the MLD pipeline we need to extract (osrm-extract) a graph out of the OpenStreetMap base map, then partition (osrm-partition) this graph recursively into cells, customize the cells (osrm-customize) by calculating routing weights for all cells, and then spawning up the development HTTP server (osrm-routed) responding to queries:


osrm-extract berlin.osm.pbf
osrm-partition berlin.osrm
osrm-customize berlin.osrm

osrm-routed --algorithm=MLD berlin.osrm 

For CH the partition and customize pipeline stage gets replaced by adding shortcuts from the Contraction Hierarchies algorithm (osrm-contract):


osrm-extract berlin-latest.osm.pbf -p profiles/car.lua
osrm-contract berlin-latest.osrm

osrm-routed berlin-latest.osrm

Extracting the Road Network (osrm-extract)

Exported OSM data files can be obtained from providers such as Geofabrik. OSM data comes in a variety of formats including XML and PBF, and contain a plethora of data. The data includes information which is irrelevant to routing, such as positions of public waste baskets. Also, the data does not conform to a hard standard and important information can be described in various ways. Thus it is necessary to extract the routing data into a normalized format. This is done by the OSRM tool named extractor. It parses the contents of the exported OSM file and writes out routing metadata.

Profiles are used during this process to determine what can be routed along, and what cannot (private roads, barriers etc.).

It's best to provide a swapfile so that the out-of-memory situations do not kill OSRM (the size depends on your map data):

fallocate -l 100G /path/to/swapfile
chmod 600 /path/to/swapfile
mkswap /path/to/swapfile
swapon /path/to/swapfile

Note: this does not write 100 GB of zeros. Instead what it does is allocating a certain amount of blocks and just setting the 'uninitialized' flag on them, returning more or less immediately.

External memory accesses are handled by the stxxl library. Although you can run the code without any special configuration you might see a warning similar to [STXXL-ERRMSG] Warning: no config file found. Given you have enough free disk space, you can happily ignore the warning or create a config file called .stxxl in the same directory where the extractor tool sits. The following is taken from the stxxl manual:

You must define the disk configuration for an STXXL program in a file named '.stxxl' that must reside in the same directory where you execute the program. You can change the default file name for the configuration file by setting the enviroment variable STXXLCFG .

Each line of the configuration file describes a disk. A disk description uses the following format:


Example: at the time of writing this (v4.9.1) the demo server uses 250GB stxxl file with the following configuration:


Check STXXL's config documentation.

Creating the Hierarchy (osrm-contract)

The so-called Hierarchy is precomputed data that enables the routing engine to find shortest path within short time.

osrm-contract map.osrm 

where map.osrm is the extracted road network. Both are generated by the previous step. A nearest-neighbor data structure and a node map are created alongside the hierarchy.

Running the Engine via a HTTP Server (osrm-routed)

We provide a demo HTTP server on top of the libosrm library doing the heavy lifting. You can start it via:

osrm-routed map.osrm

You can access the API on localhost:5000. See the Server API for details on how to use it.

Here is how to run an example query:

curl ",52.517037;13.385983,52.496891?steps=true"

Running multiple profiles on one machine

Below is an example nginx.conf configuration that shows how to configure an nginx as a reverse proxy to send different requests to different osrm-routed instances:

server {
        listen 80 default_server;
        listen [::]:80 default_server ipv6only=on;

        # catch all v5 routes and send them to OSRM
        location /route/v1/driving   { proxy_pass http://localhost:5000; }
        location /route/v1/walking   { proxy_pass http://localhost:5001; }
        location /route/v1/cycling   { proxy_pass http://localhost:5002; }

        # Everything else is a mistake and returns an error
        location /           {
          add_header Content-Type text/plain;
          return 404 'Your request is bad and you should feel bad.';

This setup assumes you have 3 different osrm-routed servers running, each using a different .osrm file set, and each running on a different HTTP port (5000, 5001, and 5002).

The driving, walking, cycling part of the URL is used by nginx to select the correct proxy backend, but after that, osrm-routed ignores it and just returns a route on whatever data that instance of osrm-routed is running with. It's up to you to ensure that the correct osrm-routed instance is using the correct datafiles so that /driving/ actually routes on a dataset generated from the car.lua profile.