Abstract: In distributed system, computing job flow is unable to be dynamically adjusted after being mapped to the nodes, which makes the essential jobs not be executed because of the inter job waiting. Aiming at the problem, this paper proposes a scheduling algorithm for computing job flow. After the jobs being mapped to distributed nodes, this algorithm gets the order values according to their dependence relations, dynamically adjusts the priority based on order values to make the essential jobs finished as soon as possible and reduce the inter job waiting, which can greatly shorten the execution time of computing job flow. The application in actual system indicates that this algorithm has strong superiority in fast execution of large number of computing job flow in job management system.

Key words: computing job flow, dependence relation, order value

摘要： 分布式系统中计算作业流被映射到节点后无法进行动态调整，使关键作业无法及时执行而造成作业间等待。针对该问题，提出一种计算作业流均衡调度算法。算法对映射到分布式节点的作业根据其依赖关系得出阶位值，依据该值在分布式节点上进行动态优先值调整，使关键作业尽早完成，减少作业之间的等待，缩短计算作业流执行时间。实际系统应用表明，该算法对作业管理系统中投入的计算作业流的快速执行有较强优越性。

关键词: 计算作业流, 依赖关系, 阶位值

CLC Number: 

• TP312