并行累积可视性分析中的任务调度研究  

Research on Task Scheduling Based on Parallel Cumulative Visibility Analysis

作　　者：宋效东[1] 窦万峰[2,3] 汤国安[1] 张刚[1] 江岭[1] 

机构地区：[1]南京师范大学虚拟地理环境教育部重点实验室,江苏南京210023 [2]南京师范大学计算机科学与技术学院,江苏南京210023 [3]江苏省信息安全保密技术工程研究中心,江苏南京210023

出　　处：《地理与地理信息科学》2013年第4期56-61,共6页Geography and Geo-Information Science

基　　金：国家863计划项目(2011AA120303);国家自然科学基金项目(41171298;41071244);江苏省普通高校研究生科研创新计划项目(CXZZ12_0393)

摘　　要：为提高并行地形分析的性能,需要对分布式并行任务进行合理的划分与调度。鉴于基于有向无环图任务划分方法存在的不足,提出了基于任务粒度模型的优化调度方法。以并行累积可视性分析算法作为用例,采用分组的调度策略实现具有不同依赖关系子任务的合并。为体现任务粒度的性能优势,使用Bresenham算法计算不同种类的可视域。在小规模集群环境的实验结果表明,基于该任务调度方法的并行算法具有较高的计算效率。To improve the performance of parallel digital terrain analysis, the computing tasks in distributed and parallel compu- ting system should be partitioned and scheduled reasonably. An improved scheduling strategy of task granularity is proposed in this paper, due to the shortcomings of tasks partitioning based on Directed Acyclic Graph （DAG）. The cumulative visibility is researched for its complexity and task dependency. For the comparison of different parallel scheduling strategy, the Bresenham method is employed to achieve the viewshed analysis. Experiments of parallel cumulative visibility analysis on small scale cluster show that the optimized partitioning method can improve the performance of viewshed analysis obviously.

关 键 词：集群 数字地形分析 并行任务调度 负载均衡 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] P208[自动化与计算机技术—计算机科学与技术]