Research On Spatial Data Clustering Services Architecture And Algorithms

The spatial data clustering services architecture and algorithms mainly for the data-intensive and high-speed required applications are studied in this paper, in which the two key points are to manage the massive spatial data and provide a high-speed data service. The article focuses on the architecture, data storage model and network distribution methods, and designs and implements a prototype named TSS. The main contents of this dissertation are as follows:1. On the basis of comprehensive analyses of the architecture of HPSS, MARS, DPSS and other clustering services systems, a new spatial data clustering service architecture is proposed, in which the structures and functions of the name server, disk server, data loader, and application programming interfaces are discussed in detail.2. A spatial data storage model for clustering service named spatial data tile model is pro-posed, which has such characteristics as direct local access, simple cache replacement, quick dis-tributed storage, and multiple optimization strategies and so on. The raster data tile model, vector data tile model and tile pyramid model are expounded. The virtual points based organization method for vector data tiles and the main data structures of the tile model are designed.3. By analyzing the necessity of the data distribution in the spatial data clustering service, the two-dimensional distribution for tiles is defined. Several distribution methods for the tile ma-trix are investigated and the performances are compared through the simulation experiments. The problem of data distribution and tiles declustering over non-uniform multi-disks is discussed. A multi-constraints considered uniform rate logical disk partition method and a maximum-distance rule based logical disk sorting method in the same physical disk or host are also given.4. A spatial data clustering services prototype-Tile Server System (TSS) is designed and implemented. Experimental results show that TSS can manage massive spatial data and provide high-speed data services. Besides, TSS also has a good trait of concurrency.