The Research And Realization Of The Partitioning Strategy Of Vector Spatial Data In Parallel Computing Environment

GIS industry has flourished over the past 50 years, promoting the rapid progress of its technology. And with the constantly expanding of geo-spatial data scale, the increasing complexity of spatial operation. traditional single GIS mode can no longer meet new requirements of the mass geo-spatial data operation, on the other hand, as the computer hardware is greatly improved, high performance parallel GIS was born. How to maximize parallel equipment computational ability are the hot research topic, the parallel task division and geo-spatial data partition are put on the strategy study as the precondition of GIS further performance. Determined the needs of parallel computing GIS, the paper study the division of space-based aggregation strategy, further study the partition method based on space cluster and put forward a new one based on space gathering, then a standard came up which will guarantee partition results, that is well space proximity, balanced data load, less data redundancy and low partition time scale.Based on the development of technology and the practical problems of GIS applications, the paper discuss algorithms and strategy for vector spatial data partitioning according to the characteristics of spatial data, combined with advances in computer science and technology. And Experiment is conducted in parallel environment deployment and tested. The paper reviews the research and development in related fields including cloud computing and parallel computing, comprehensively analysis the current research development and application of cloud computing and data at home and abroad in a thorough understanding of its core technology, which provide a theoretical basis and technical support for the further study. Also, the direction and key technique of the paper are established which is the methods and strategies for vector spatial data partitioning taking into account the shapes of spatial data.The paper introduces the commonly used data partitioning method, and defines the principle of data partitioning according the characteristics of vector spatial data, there are data partitioning load metrics, the location of the spatial vector object, no intersect principle of spatial data, spatial proximity. The problems of the traditional method for vector spatial data partitioning are analyzed. And the key problem of data partitioning is described in parallel environment, which is a set of spatial operations to a spatial data set on N-storage space, assigning different storage strategies for fully using storage resources to improve the response time of spatial operations through effective data partitioning.Paper focuses on vector spatial data partitioning methods and strategy in the parallel environment. According to the core mission of spatial data partitioning and the features of spatial data aggregation and statistical clustering, the paper designs two strategies. The nature of spatial data partitioning is dividing a data set into several sub-blocks based on the aggregation level of spatial. The feature of spatial aggregation mapping spatial data from multi-dimensional set to one-dimensional linear meets that demand. Peano in space-filling good curve family has the spatial ability of spatial clustering through traversing all spatial objects one time. Curve partitioning algorithm represented by the Hilbert creates a good spatial arrangement of the object for spatial partitioning coding by the order code, which are high efficiency, low time consumption ratio and high data accuracy. Based on statistical clustering, a new data partitioning method is designed, classifying spatial objects as the same class with the same property in space. With similar spatial characteristics of the statistical data as a criterion, spatial objects are clustering. In order to ensure proximity of spatial data, the nearest cluster or area of the smallest increment of the standard is used to get a better partitioning result.To verify the validity and feasibility of the key methods and techniques in the paper, the experiment is completed in the parallel environment for different types of spatial elements (point, line, surface) and numbers of sets. Universal and cover the general characteristics of the data (uneven distribution of data, massive amounts of data, etc.) experimental data are selected for the two splitting strategy to test accuracy and efficiency. Using the actual operation efficiency and the optimal use of computing resources as the ultimate goal, the paper proposes the dynamic selecting strategy for spatial data partitioning in different, different forms of data, different data distribution and different computing requirements. This study that methods and strategies for the vector spatial data partitioning can be applied to spatial analysis in various types of parallel computing environment and distributed disposition of vector spatial data.