动态多DAG调度的改进算法  

An improved algorithm for dynamic multi-DAG schedule

作　　者：张明会[1] 周勇[2] 赵新政[2] 

机构地区：[1]大连东软信息学院计算机科学与技术系,辽宁大连116023 [2]大连理工大学软件学院,辽宁大连116620

出　　处：《中国科技论文》2015年第14期1651-1655,共5页China Sciencepaper

基　　金：高等学校博士学科点专项科研基金资助项目(20120041120052)

摘　　要：工作流调度问题通常可以描述为一个活动与数据传输约束组成的有向无环图(directed acyclic graph,DAG),基于Planner-guided算法框架,采用新的基于多有向无环图节点合并的优先级计算方案改进原有算法,改善了系统中已存在未执行的DAG结点优先级过小而被滞后的问题,从而使得任务在机器上排布相对比较密集,提升了多个DAG结点间调度公平性。结果表明,提出的改进算法优先级计算更加合理,在资源利用率和执行时间上优于Planner-guided算法,调度整体性能得到一定提升。Workflow scheduling problem is usually modeled as a directed acyclic graph (DAG)comprised by an activity and data transmission constraints.A new priority setting algorithm is proposed based on the architecture of Planned-guided algorithm. This algorithm improves the delay problem of the node which has not been executed because of its low priority,and makes the job scatter relatively intensively.In the end the performances are analyzed and experiment results show that the algorithm outper-forms Planner-guided algorithm in utilization of machine resources and execution time.

关 键 词：工作流调度 有向无环图 优先级计算 资源利用率 

分 类 号：O157.5[理学—数学]