Research On Self-healing Key Technology Of Geospatial Service Composition In Cloud Environment

As an expression and implementation of the complex geoprocessing model in the distributed computing environment, Geospatial Service Composition (GSC) is the major research direction in the geospatial service oriented computing realm. Self-healing mechanism is the ability for self-monitoring, self-diagnosing and self-recoverying when the system encounters failure. With the migration of geocomputing environment from desktop to cloud computing, the self-healing demands in GSCs become urgent. The fundamental problems are how to describe self-healing action in GSC models and how to satisfy the accurate and real-time self-healing needs in cloud environment that make geospatial services compositions achieve ubiquity.In this thesis, the failure mechanism in the GSCs in the cloud environment is studied and analyzed in depth. Based on this, the failure types and recovery actions is mapped and extended into the different cloud computation levels. A self-healing policy model is built systematicly. Then spatio-temporal features optimized geospatial services substitution and GSC reconfiguration algorithms are proposed in this paper. Finally, a self-healing framework prototype on Azure Cloud is set up. The main contents are listed below:1) Geospatial service compositions expression and formal description methods. Learning from the advantages and disadvantages in the existing failure types and recovery actions, this paper systematically analyses the failure classification and recovery methods in the cloud environment. A self-healing policy description model for GSC called SHPolicy is proposed. It is the model foundation for self-healing policy design, and also the theoretical basis and guidance for the running self-healing mechanism.2) Geospatial services substitution oriented QoGIS prediction method. In this paper, a deep analysis is taken in the spatio-temporal features influences on QoGIS. Based on this, a QoGIS prediction method with spatio-temporal features is designed. The timezone and timeslice are considered as the spatio-temporal features of the QoGIS. The differences of the spatio-temporal features between the client and server are taken a full account. A Space-Time Aware Collaborative Filtering (STACF) algorithm for QoGIS prediction is proposed with the spatiol-temporal features involving the computation procedure. It aims to improve the prediction accuration of the QoGIS prediction method. The feasibility is proved by analogue experiments. Comparison with similar algorithms shows that the STACF algorithm is more accurate and effective.3) The spatio-temporal features optimized GSC reconfiguration algorithm. Learning from the current GSC reconfiguration algorithm, the spatiol-temporal constraints features for GSCs are defined firstly. Then the skyline dominance degree technology is used to filter the geospatial services candidates based on the traditional genetic algorithm. A space aggregation degree concept is defined. And a fitness function based on the space aggregation degree is designed to improve the quality of the approximate optimal solution and control the dispersion level among the atomic geospatial services. Based on this, a Space-Time Optimized Geospatial Service Composition Reconfiguration (STOR) algorithm is proposed. It takes a full consideration of spatial-temporal features'influences on QoGIS prediction. It optimizes the initial population and services candidates in order to improve the convergence speed. The STOR algorithm is an available and effective solution for the GSC reconfiguration problem.4) The prototype of the GSC self-healing framework in cloud environment. A theoriotical framework model in cloud computing is designed. The corresponding physical model is implemented on Azure cloud. It integrates the SHPolicy model, STACF algorithm and STOR algorithm. A completed self-healing solution for GSCs in real cloud platform is set up.The research results from this study will help the research on the robustness of GSCs in complex and dynamic cloud environment. That is exploration from "how can describe the self-healing actions?" to "how to apply the spatial-temporal features optimizing recovery algorithms to build an effective available collaborative geoprocessing model, and to improve the utilization of the cloud computing resources? ".