Geowave: Utilizing distributed key-value stores for multidimensional data
MA Whitby, R Fecher, C Bennight - … SSTD 2017, Arlington, VA, USA, August …, 2017 - Springer
MA Whitby, R Fecher, C Bennight
Advances in Spatial and Temporal Databases: 15th International Symposium, SSTD …, 2017•SpringerTo date, it has been difficult for modern geospatial software projects to take advantage of the
benefits provided by distributed computing frameworks due to the implicit challenges of
spatial and spatiotemporal data. Chief among these issues is preserving locality between
multidimensional objects and the single dimensional sort order imposed by key-value
stores. We will use the open source framework GeoWave to harness the scalability of
various distributed frameworks and integrate them with geospatial queries, analytics, and …
benefits provided by distributed computing frameworks due to the implicit challenges of
spatial and spatiotemporal data. Chief among these issues is preserving locality between
multidimensional objects and the single dimensional sort order imposed by key-value
stores. We will use the open source framework GeoWave to harness the scalability of
various distributed frameworks and integrate them with geospatial queries, analytics, and …
Abstract
To date, it has been difficult for modern geospatial software projects to take advantage of the benefits provided by distributed computing frameworks due to the implicit challenges of spatial and spatiotemporal data. Chief among these issues is preserving locality between multidimensional objects and the single dimensional sort order imposed by key-value stores. We will use the open source framework GeoWave to harness the scalability of various distributed frameworks and integrate them with geospatial queries, analytics, and map rendering. GeoWave performs dimensionality reduction by utilizing space–filling curves to convert n-dimensional data into a single dimension. This ensures that values close in multidimensional space are highly contiguous in the single dimensional keys of the datastore. By using various forms of geospatial data, we show that preserving locality in this way reduces the time needed to query, analyze, and render large amounts of data by multiple orders of magnitude.
Springer