Research Of Service Architecture And Algorithms For Grid Spatial Data

In this thesis, the service architecture and algorithms for grid spatial data are researched to design a high-performance Terrain Server System (TSS) which manages massive data and provides high-speed data service. The grid spatial data refers to terrain data and ground surface image that include multi-source Remote Sensing Image, Digital Elevation Model and Digital Ortho Image. The dissertation is centered on TSS architecture, data management models, data declustering methods and server design. Finally, a prototype system has been designed and realized. The works of this thesis can be summarized as follows:1. A new distributed-parallel architecture of TSS is proposed based on analysis of the architecture of high performance storage system such as HPSS, MARS, SAN and DPSS. The architecture and functions of each component of TSS are introduced in detail.2. Tile-pyramid Model and Linear Quadtree Tile-index are put forward for applications of massive terrain data and data declustering on multi-disks. More emphasis is given to study the rule of data blocking and common method of building tile-pyramid model and linear quadtree tile-index. On the other hand, this thesis also investigates such problems as target-tile searching, tile-request prediction and edge matching of block terrain drawing.3. The necessity of applying data declustering in TSS is discussed and two-dimensional declustering for tile-matrix of terrain data is defined. Several methods for two-dimensional data declustering are discussed and their performances are evaluated through experiments. After studying the questions of data allocation and tile declustering in heterogeneous disk systems, a same-rate virtual disk dividing method and a virtual disk sorting method are proposed.4. A multi-thread and parallel server architecture for high-performance terrain data service is put forward based on analysis of advantages and disadvantages of typical models for designing server.5. A prototype system TSS has been designed and realized based on gigabit-per-second LAN. Experimental results show that the prototype system can manage massive terrain data and has good concurrent performance. Its average throughput is 5.6 times than that of single disk server under same conditions.