Studying On The Technolgies Of Storage And Processing Of Spatial Vector Data In High-Performance Parallel GIS

Traditionally, the geo-spatial data information can be understood as the blood of geographical information System-GIS in the field of Geo-spatial Information Science, and GIS plays an important role in organizing and management of a huge volume of geo-spatial data. Currently, with the development and advancement of earth observation system, the volume of geo-spatial data we can harvest is in the blowing up increase. As a result, GIS will unavoidably face to the more challenges that the inevitable development trend brought up, data, information, knowledge, and the more application requirements will be submitted to GIS itself. However, as an advanced drive, the development of the theory and technologies of GIS will go ahead toward to the highest objective.WebGIS and distributed GIS that the integrating result of variant computing schema and GIS implement primarily the geo-spatial information services based on communication network, but the existing technologies have not resolved the fundamental problems to which the massive geo-spatial information services faced. Recently, grid computing becomes a research hotspot in the field of information technology science, and it has the huge potential power in geo-spatial information services. Essentially, the introducing of grid concept and the integrating of grid computing technologies and GIS is the other renovation of computing schema, and its objective is to conform the various resources, such as computing resource, storage resource, equipment resource, and so on, and implement the distributed processing and services based on network of the huge volume of geo-spatial information. However, for implementing the objective, the high performance storage, retrieval and processing of the huge volume of geo-spatial data is indispensable precondition. Therefore, the research interesting of the paper that the key technologies of high performance storage, retrieval and processing of vector geo-spatial data in high performance parallel GIS have practical and applied value.The paper follows the traditional research method of the theory and the technologies of GIS, such as parallel geo-spatial data structure, parallel geo-spatial index mechanism, parallel geo-spatial operation and parallel geo-spatial applications etc., and analyses and address deeply the basic theory and the key technologies of the high performance GIS and parallel geo-spatial database system. On the aspect of the design of parallel geo-spatial data structure, the paper focuses on the modeling methods of vector geo-spatial data structure, and follows the specification of the simple feature modeling provided by OGC organization and describes the data structure using object-oriented programming. The modeling methods of vector geo-spatial data based on object-oriented theory have the inherent characters of parallelizability, and are propitious to implement parallel I/O operation of parallel geo-spatial database in order to improve the capability of parallel processing of geo-spatial data. The paper builds the shared-nothing structure parallel geo-spatial database based on object-relation database system act as the basic database platform; the vector data and attribute data would be stored and managed uniformly in the same relation table. In other words, it means that the huge volume of the spatial data should be corresponding to the massive data items in geo-spatial database system. Thus, the policy of parallel partitioning of geo-spatial data will be one of the important factor that effect the efficiency of parallel geo-spatial database system. The paper presents a new parallel partitioning policy of the geo-spatial data based on the Hilbert spatial filling curve, HCSDP. Different from those existing data partitioning policies, the characteristics of the geo-spatial data, such as the aggregation of geo-spatial data, can be considered fully by HCSDP and the storage load among the different spatial sub-regions partitioned by HCSDP will be balanced, it ensures the static storage load balancing of parallel geo-spatial database. Geo-spatial index mechanism is the key issue for the data organization and data management in GIS. To parallel GIS, the parallel index mechanism will play an very important role for improving the capability of storage and management of the huge volume of geo-spatial data in parallel GIS and parallel geo-spatial database system. The paper presents a new multi-level parallel R-tree geo-spatial index structure based on the Hilbert space filling curve HCMPR-tree, and gives the design and building method of HCMPR-tree. Comparing to the existing parallel geo-spatial index mechanisms, HCMPR-tree avoids the possibility that the main controlling computing nodes in the network of the PC become the hotspot. The well structure's characteristics of HCMPR-tree support appropriately the design and the implementation of the basic algorithms of parallel geo-spatial operation and parallel applications addressed in the paper. As the basal functions of the geo-spatial database system, the paper gives the parallel design and parallel implementation of the two basal geo-spatial operations in the parallel geo-spatial database system, that is, the parallel geo-spatial range query and the parallel geo-spatial join. The parallel design and parallel implementation of the two geo-spatial operations is based on the index structure of the HCMPR-tree, and the complexity and the efficiency of the two algorithms will be depended on the characteristics of HCMPR-tree. In the paper, a cost model of the algorithm of the parallel geo-spatial operations is put forward in order to evaluate quantitatively the performance efficiency of these algorithms. On the aspect of the research on the parallel geo-spatial applications of parallel GIS in the paper, firstly, the paper gives the generic theory and process of the parallel geo-spatial applications supported by parallel GIS and parallel geo-spatial database system combining with the application instance of the 3-dimension DEM visualization. In the section of the paper, the research work focuses on the problem of the optimization path analysis in the geo-spatial network analysis, and gives the design and the implementation of the parallel optimization path analysis algorithm based on parallel GIS and parallel geo-spatial database system and addresses deeply those key issues. To the design of parallel optimization path analysis algorithm, the organization of the huge complicated volume of the geo-spatial network data and the technologies of graph partitioning is the key problems. Therefore, the paper adopts the graph partitioning policy based on HCMPR-tree index structure, and gives the evaluation standard of the efficiency of the graph partitioning policy and its implementation. Finally, the paper gives the design and implementation of the parallel optimization path algorithm based on the classical Dijkstra algorithm running on the parallel computing environment of the network of PC, and proves the validity and the feasibility of the parallel geo-spatial applications supported by parallel GIS and parallel geo-spatial database system presented in the paper.The theory methods and the key technologies adopted in the paper is propitious to form a suit of more integrated research infrastructure in order to direct the research and practice work of high performance parallel GIS and parallel geo-spatial database system.