The article analyses the behavior of two kinds of multiprocessor multidisk storage server architectures for a data intensive application, namely for spatial queries in geographical information systems (GIS). The two kinds of servers are: (1) a workstation cluster architecture with multiple processors, multiple disks, and a shared bus shared memory architecture; (2) a distributed memory architecture, similar to the T9000 transputer based architectures, where processing nodes and disk nodes are connected by a high throughput crossbar switch. The GIS application under investigation is the map overlay with map at different scales and resolutions
Rachid Guerraoui, Mihail Igor Zablotchi, Julien Michel Stainer, Peva François Blanchard