A Quantitative Analysis of a

Introduction The current trend in geographic information systems (GIS) is to move away from standalone systems to those that are distributed and can provide processing capability across a network. Additionally, GIS technology is finding its way into more applications and becoming a more ubiquitous technology. Many times, users are not even aware that they are using GIS technology. As these technology trends occur, the traditional GIS analyst is using data sources he/she previously did not use in the analytical process. Examples include image, video, audio, and textual data. To address these trends, we present an agent-based approach to a distributed GIS. The approach outlines a methodology for taking legacy, stand-alone applications and representing them as a set of distributed collaborative agents. The architecture is demonstrated though several different classes of agents that all serve the GIS analyst. These agents perform GIS analysis, image processing, natural language text processing, data mining, and websearching. We report on the overall agent-architecture: the granularity of the agents, the ontology, and the agent communication language. We present results on the cost of integrating disparate legacy software packages, an agent-discovery mechanism, and an agent mobility model that migrates agents based on the size of the job, as well as the state of processors on the network.