基于空间聚类的矢量空间数据并行计算划分方法

发布时间：2018-04-18 07:47

本文选题：并行计算 + Hilbert曲线　；参考：《高技术通讯》2015年04期

【摘要】：为了解决并行矢量空间分析在数据划分阶段的负载均衡问题,研究了矢量空间数据的划分,提出了一种基于空间聚类思想的矢量空间数据划分方法。该方法充分考虑矢量空间数据规模以及空间邻近性特征对并行空间分析算法效率的影响,首先采用空间填充曲线对二维空间数据进行编码,保证空间要素邻近性特征;然后用空间要素集合对空间要素流进行填充,从而确保各个子任务集中的要素数据规模相对均衡。以并行叠加分析中点面、线面、面面叠加操作为例,设计了对比实验。实验结果表明,该方法能够有效提高以线、面要素为操作对象的并行算法负载均衡度和提高并行算法整体运行效率。
[Abstract]:In order to solve the problem of load balancing in the phase of data partition in parallel vector space analysis, the partition of vector space data is studied, and a method of vector space data partition based on spatial clustering is proposed.The effect of vector spatial data size and spatial proximity feature on the efficiency of parallel spatial analysis algorithm is fully considered in this method. Firstly, the spatial filling curve is used to encode two-dimensional spatial data to ensure the spatial element proximity feature.Then the spatial element set is used to fill the spatial element flow, so as to ensure that the scale of the element data in each subtask set is relatively balanced.Taking the operation of point surface, line surface and surface superposition for example, a comparative experiment is designed.Experimental results show that the proposed method can effectively improve the load balancing degree of parallel algorithm with line and surface elements as the operating object and improve the overall running efficiency of the parallel algorithm.
【作者单位】：中国科学院计算技术研究所;中国科学院大学;香港理工大学;
【基金】：863计划(2011AA120302) 国家青年科学基金(41301423)资助项目
【分类号】：TP338.6

【参考文献】