面向并行空间连接的两轮映射数据划分方法

摘要

Too much repeated spatial objects (the redundant data) which are contained by the partitioned data blocks and low-level data balance degree among the partitioned data blocks would increase the system's work load and reduce the system's load balance.The two-rounds-map partitioning method was proposed to solve the problem.Redundant data was reduced by making full use of the space attribute of the partitioned objects within the first round map,and spatial data was evenly partitioned by reasonable setting threshold.Data balance degree was improved by dynamic mapping mechanism within the second round map.Experimental results show that the two-rounds-map partitioning method can effectively reduce redundant data and dramatically improve the data balance degree of partitioned results compared with Oracle Spatial,Linear-Round-Robin and Hilbert-Round-Robin.The two-rounds-map partitioning method has a good computational efficiency.%针对数据划分结果高冗余、低均衡可能会增加系统的工作负荷和影响系统的负载均衡这一问题,提出两轮映射数据划分方法.在第一轮映射中,通过充分利用划分对象的空间属性来减少冗余数据,通过合理设置阈值来均衡划分数据;在第二轮映射中,通过动态映射机制,提高划分结果的数据量均衡度.与Oracle Spatial数据划分方法、线性编码轮询调度划分方法以及Hilbert编码轮询调度划分方法进行比较可知,采用两轮映射方法可以有效地控制冗余数据的产生,大幅提高划分结果的数据量均衡度,具备较好的划分效率.