Spatial Query Performance Analyses on a Big Taxi Trip Origin–Destination Dataset

Abstract

Database management systems (DBMSs) are widely used to store, query and visualise large amounts of spatial data. Consequently, teaching DBMSs at both undergraduate and graduate level is a core component within most of the geoinformatics related departments’ curricula. Traditional teaching material mainly covers relational DBMS, due to their strong linkage with GIS. However, due to the emergence of web-based systems, employers also require prospective graduates to have experience on non-relational DBMS (NoSQL). This chapter develops an open-source library to analyse the query performance of two renowned DBMS: PostgreSQL/PostGIS and MongoDB, a relational and NoSQL DBMS respectively. An open-source Python library is developed to facilitate systematic performance analyses between these DBMSs. The experiments are carried out on New York City’s openly available taxi trip origin–destination dataset. The performance of two spatial queries (k-nearest neighbour and point-in-polygon) are investigated in terms of run-time and spatial accuracy. The results indicate the superiority of MongoDB. It outperformed Postgres in terms of run-time in both of the investigated queries. In addition, it is more accurate in terms of detecting k-nearest neighbours. The developed open-source library is utilised to investigate journey time variations between two airports of New York City, which demonstrates its effectiveness in terms of teaching DBMS or GIS modules.