Abstract
Geoprocessing on a High Performance Cluster
Track: Analysis
Authors: Benjamin Mearns
Researchers at the University of Delaware (UD) are often faced with time-consuming geoprocessing of large datasets. With recent acquisition and deployment of our > 5,000 core computing cluster, Mills, we saw an opportunity to considerably reduce the time cost in geospatial computing. ArcGIS deployment for high performance multicore geoprocessing on Linux systems, particularly on clusters, presents distinct challenges. This talk will cover our experiences and details of our implementation on Mills, which addresses our most commonly requested and most time consuming compute needs.