Data retrieval for small spatial regions in OpenStreetMap

Abstract

Using OpenStreetMap data usually means, firstly, filtering for thematic extracts. This can be done with the OpenStreetMap mirror database Overpass API (application programming interface). To do this efficiently, the database takes advantage of the assumption that data is often selected from a relatively small spatial region instead of randomly across the planet. The paper aims to investigate what design choices are required to be able to answer almost any geographic query whilst serving common use cases fast enough such that the services based on this database are fast on affordable and standard sized hardware. The usage patterns from the main instance of Overpass API on overpass-api.de are evaluated. These comprise more than 40 million requests from the years 2012 to 2013 coming from about 60 % of the global IPv4 (IP protocol version 4) space. Therefore it can be said for sure that the assumption that queries are spatially dense holds on a large share of all queries.