Benchmarking the Performance of a Cluster-Based Geospatial Database System

The advent of inexpensive high performance computing clusters has opened the doorway to a wide range of applications that hitherto had been inconceivable. Such clusters can provide certain applications with a substantial boost in both compute and I/O performance. Therefore, it seems logical that Geographic Information Systems (GIS) applications would be an ideal fit for such systems. This paper presents a scheme to distribute geospatial data across several compute nodes in a cluster. Several variations of this scheme are presented and the performance of each is benchmarked by using various geospatial queries to retrieve the data. The original document contains color images. All DTIC reproductions will be in black and white. Prepared in collaboration with Dept. of Computer Science, Univ. of New Orleans, New Orleans, LA.