Backend for an OSM data analysis tool. Uses osm-qa-tiles as input data.

If you want to use the cruncher natively, you'll need:

• NodeJS
• NPM

Be sure to run npm install before running the commands bellow

Alternatively, you can install and use docker and docker-compose.

The tool is split in two parts: data generation and serving.

Invoking app/run.sh (or ./docker.sh gen command in a docker environment) starts the process to re-generate osm-analytics data. It will automatically download a fresh osm-qa-tiles planet file, crunch the data and stores the results in a supplied directory.

The data generation can be configured by setting the following shell environment variables:

• ANALYTICS_FILE – file defining analytics job (e.g. what layers to crunch), see example-analytics.json for an example
• WORKING_DIR – working directory where intermediate data is stored
• RESULTS_DIR – directory where resulting .mbtiles files are stored
• OSMQATILES_SOURCE – URL to fetch osmqatiles from

The data generation process is controlled via a single "analytics definition file", which specifies which parts of the OpenStreetMap data has to be processed in which way and how the result should be interpreted. An example can be found at app/example-analytics.json. Further details about this file can be found in the specification document.

The app/server/serve.js script is an example for how to provide the data to the osm-analytics frontend over the web. It expects the directory where the result of the crunching step is stored and the analytics definition json file as parameters.

You can use the ./docker server buildings.mbtile command to serve a pre-computed mbtile file. The mbtile file must be available in the ./results local folder.

The cruncher alternates between periods of heavy processing and no load at all after that. Due to this, it is worth considering an execution environment that does not rely on an always-on server, but rather on a job-based approach. This can be achieved using Google Cloud and the following command:

gcloud compute --project "osma-174310" instances create "osma-cruncher" --zone "us-central1-c" --machine-type "custom-16-30720" --subnet "default" --maintenance-policy "MIGRATE" --service-account <your google data> --scopes "https://www.googleapis.com/auth/cloud-platform" --image "osma-cruncher-v2" --image-project "osma-174310" --boot-disk-size "100" --boot-disk-type "pd-ssd" --boot-disk-device-name "osma-cruncher"

Its worth noting that this is not a requirement of the cruncher, nor a dependency on Google Cloud. A similar approach can be achieved with different hosting services with no code changes, and the cruncher can be set up periodically on a "traditional", always-on server.

The crunching process is a resource-intensive task, requiring significant CPU, memory and storage to execute. To the date of these tests, the cruncher required 16 processing cores, 30GB of RAM and 100GB of storage to process the data in around 6h.

The necessary storage space depends on the amount of data input and output, so it may need to vary if the OSM tiles grow in size, or if more features need to be processed.

CPU and RAM are linearly correlated, and mostly dictate the time it takes to process the whole data.

An overview of all steps required to implement an instance of osm-analytics can be found here (work in progress).

A schematic diagram of the different components of the cruncher are found in the documentation directory.