Monet and its Geographical Extensions: a Novel Approach to High-Performance GIS Processing

We describe Monet, a novel database system, designed to get maximum performance out of today's workstations and symmetric multiprocessors. Monet is a type- and algebra-extensible database system using the Decomposed Storage Model (DSM) and employing shared memory parallelism. It applies purely main-memory algorithms for processing and uses OS virtual memory primitives for handling large data. Monet provides many options in memory management and virtual-memory clustering strategies to optimize access to its tables. We discuss how these unusual features impacted the design, implementation and performance of a set of GIS extension modules, that can be loaded at runtime in Monet, to obtain a functional complete GIS server. The validity of our approach is shown by excellent performance figures on both the Regional and National Sequoia storage benchmark.