Research on Parallel Geo-spatial Index Replication Strategy Based on PC Cluster System and its 3D Visualization Application

摘要

In the field of spatial information, the amount of spatial data becomes more and more large, and every operation becomes more and more complicated. Now, parallel techniques gradually become valid means of resolving this kind of complicated problem. Therefore, this paper studies the spatial partitioning method of massive data and the parallel geo-spatial index replication strategy after fully considering characteristics of PC cluster based on shared-nothing structure. After studying excellent linear mapping characteristics of Hilbert spatial ordering code, this paper applies it to spatial partitioning of data, and gives a concrete algorithm. In this algorithm, the clustering performance of spatial objects is considered, and the balance of data storage on each processing unit is also done, which greatly improves the processing efficiency of parallel spatial database. Based on this, this paper builds the parallel geo-spatial index based on R-tree, and proposes the spatial index replicas update mechanism based on master replica and weak consistency suitable for the parallel environment of shared-nothing structure after deeply analyzing current synchronous mechanisms of replicas. It is a kind of update mechanism based on message with low cost, enhances the power of parallel spatial data access and using PC cluster, and improves the availability of index data. In the field of 3D visualization application, it is also efficient. Experiments prove that the spatial partitioning strategy and the parallel geo-spatial index replication mechanism presented in this paper can improve load balance of system, and enhance performance of the whole system.