Operator scheduling in a distributed stream management system for remotely sensed imagery

Large amounts of remotely sensed geospatial image data are continuously streamed to Earth fro satellites. Many applications, such as environmental monitoring, climatology, and disaster management use the raster images obtained from satellites.;In this dissertation, we investigate distributed geospatial image stream processing systems (DGMS) with a focus on operator scheduling. In this context, we present a data and query model for geospatial image data streams (GeoStreams). The data model captures the temporal and geospatial characteristics of GeoStreams. The query model provides a set of image processing operators that allow the users to query GeoStreams. Based on the data and query model, we develop a cost model to estimate the resource requirements of the operators presented in the query model.;Operator scheduling is key to the performance of a DGMS. We show by a reduction from the bin-packing problem that the DGMS operator scheduling problem is NP-Complete. This insight into the operator scheduling problem is essential to enable the development of heuristic-based scheduling algoritlinis for a DGMS. Furthermore, we identify a new class of polynomial time heuristic scheduling algorithms referred to as spatially aware scheduling algorithms in this dissertation. Spatially aware scheduling algorithms exploit the spatial characteristics of the data streams and continuous queries over such streams. These algorithms reduce the resource consumption and minimize processing latencies when compared to a naive scheduling algorithm.;We also develop a distributed GeoStreams simulator (DiGSim). DiGSim simulates a distributed geospatial stream processing system and provides a framework to design and test the performance of DGMS scheduling algorithms.