Spatial Join Query Processing in Cloud: Analyzing Design Choices and Performance Comparisons

Data volumes of GPS recorded locations and many other types of geospatial data are fast increasing. Processing large-scale spatial joins in Cloud for performance and scalability is becoming increasingly popular. In this study, we compare three leading Cloud-based spatial data management systems, namely HadoopGIS, SpatialHadoop and SpatialSpark, both conceptually through analysis of design choices and empirically through experiments using real world datasets. Using both a workstation serving as a single-node cluster and up to 10 nodes Amazon EC2 clusters, the results show that the combined factors, including Cloud platforms, data access models and the underlying geometry libraries, have significant impacts in their realized performance. While SpatialHadoop generally wins on robustness, SpatialSpark is the clear winner of efficiency due to in-memory processing.