Loading...
Please wait, while we are loading the content...
Similar Documents
Parallel Neighbourhood Modeling: Research Summary Y 2 System Overview 2.1 Design Issues
| Content Provider | Semantic Scholar |
|---|---|
| Author | Maheshwari, Aakash Nussbaum, Daniel Roytenberg, David |
| Copyright Year | 1996 |
| Abstract | 1 Motivation and Applications In recent years the demand on Geographical Information System (GIS) technology from application areas has sharply increased. In particular, with the EOS (Earth Observation System), terabytes of spatial information will be collected daily. The increase in speed in sequential computing cannot keep up with the demand placed by many systems handling spatial data. Therefore a need for parallel computing in this area is widely recognized. Raster processing, such as neighbourhood model-ing or computing complex visibility information is, in many cases, a time consuming task. At a processing rate of 1000 cells/sec., processing a large raster of size 6000x6000 cells takes 10 hours. The problem becomes more acute when cellular automata modeling is considered. The time to process even a small number of generations , say 200, on a 6000x6000 cell raster takes over 83 days. Related problems include spatial feature extractions where \numerous investigators report lengthy computation time" 9]. Many users who execute such time intensive or time sensitive tasks are forced to either reduce the raster size, possibly sacriicing resolution and accuracy, or choose a simpler model which may sacriice strength, scope and validity. Raster process-ing/modeling operations are employed in many application areas such as remote sensing, urban planning, and simulating biological/ medical/hydrological processes. We are in the third year of a project funded in part by industry and the Canadian Government (NSERC) to develop a parallel GIS. The funding for the project is approximately $4,000,000 (CAN). Here, we describe our rst major milestone, the design and features of a parallel system for NEighbourhood MOdeling (called NEMO) of raster data. The mandate of our system is to provide users with a fast engine for processing a variety of time-consuming tasks for raster based data. The system is designed to be platform independent and is currently implemented on the AVX series II parallel computer manufactured by Alex Informatique, a Cana-dian supercomputer manufacturer. The Alex AVX Series II computer running under the Trollius operating system, is a exible distributed memory MIMD parallel machine which can be reconngured into a variety of standard and non-standard topologies. The AVX has 64 standard nodes and several entry (external access) nodes. A standard node consists of two processors, one Intel i860 (used for computations) with between 32 and 64MB of DRAM, and one Inmos T805 transputer (used primarily for communication) with between 8 and 16MB of DRAM. While parallel computers are … |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.scs.carleton.ca/~lanthier/research/SPAA96.ps |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Summary |
