Title: Distributing Computations Among GIS Servers
Author(s): M.A. Gennert, N.I. Hachem, N. Serrao, and A. Bansal, Computer Science Department, Worcester Polytechnic Institute, Worcester, MA 01609
Source: Proc. 7th Int. Conf. Parallel and Distributed Computing Systems, pp.294-301, Las Vegas, NV, Oct. 1994.
Abstract: Scientific databases, including Geographic Information Systems (GIS), pose a number of challenging problems for system architects and implementors. Foremost is the amount of data to be stored and processed, necessitating distribution of the data at multiple sites and in multiple formats to be used by multiple analysis packages. However, the end user does not care about the complexities that result, preferring to operate in a seamless environment providing transparent access to a variety of data and analysis resources.
In this paper, we present an approach to distributing computations in a GIS based on the client/server model of computation. All GIS operations are viewed as queries. If the requested data exist, they are retrieved directly. Otherwise, a set of operators capable of deriving the requested data from known data is generated and executed. Each database and analysis operator package is placed in a server which responds to requests from query/analysis processor clients. We give an example of how distributed computations can be managed using a Petri Net representation of metadata.
email@example.com / Aug 8, 1996