A parallel computing strategy named no join-general parallel computing architecture(NJ-GPCA)is proposed to solve the spatial overlay analyzing technique's drawbacks of poor I/O performance and low parallel task schedu-ling efficiency.Firstly,the strategy puts the vector spatial data into the database of Redis with the new structure, and then cuts down the processes wait time and improves the efficiency of I /O by data pre-processing and task dis-tributing.Finally,the strategy uses a new task allocation and task scheduling to avoid result collection,which re-duces the time complexity of the algorithm's collection stage to O(n)from O(nlogn).The experimental result shows that the strategy can reduce the I/O time by at least 75%and significantly improve the efficiency of the algo-rithm.%针对并行矢量空间叠加分析中存在的I/O性能差及并行算法调度效率低的缺陷,提出了"去"归并通用并行计算架构(NJ-GPCA).该架构首先基于内存数据库Redis设计内存矢量空间数据模型;其次通过数据预处理以及任务分发技术,减少进程等待,提高I/O性能;最后重新进行任务分配以及规划进程调度,避免结果数据归并收集,使得并行叠加分析算法归并收集阶段的时间复杂度由O(nlogn)降低到O(n).实验结果表明,该方法对真实地理数据下的并行叠加分析操作,I/O时间至少减少75%,对于提高算法整体性能有明显效果.
展开▼
