• 论文 •    

分布式工作流中的自适应实时任务调度方法

谭宜勇,王锐,范玉顺,郑邦东,郭关飞,孙刚,宋耀华   

  1. 1.清华大学 自动化系,北京100084;2.中国建设银行 信息技术管理部武汉开发中心,湖北武汉430015;3.京北方科技股份有限公司,北京100080
  • 出版日期:2010-09-15 发布日期:2010-09-25

Adaptive scheduling method for real-time tasks in distributed workflow

TAN Yi-yong, WANG Rui, FAN Yu-shun, ZHENG Bang-dong, GUO Guan-fei, SUN Gang, SONG Yao-hua   

  1. 1.Department of Automation, Tsinghua University, Beijing 100084, China;2.Department of IT Management, Wuhan Development Center, China Construction Bank, Wuhan 430015, China;3.NorthKing Technology Co., Ltd., Beijing 100080, China
  • Online:2010-09-15 Published:2010-09-25

摘要: 为提高分布式工作流中实时任务的调度效率,提出了一种基于比例—积分—微分控制的自适应任务调度方法。该方法采用基于动态关键路径的任务分配策略把任务分配到不同的流程引擎上,并通过任务调度器对任务进行调度。任务调度器根据准入控制器和比例—积分—微分控制器的计算结果,采用弹簧算法调整流程引擎中实时任务的各项参数,以便让流程引擎处理尽可能多的任务。此外,为了防止系统的瞬时过载,自适应任务调度方法采用了准入控制策略。该策略根据系统当前的负载情况实时对进入系统的任务数量进行控制,保证了系统的稳定运行。

关键词: 分布式工作流, 自适应调度, 比例—积分—微分控制, 关键路径

Abstract: To improve the scheduling efficiency of real-time tasks in a distributed workflow systems, an Adaptive Scheduling Method(ASM) based on Proportional-Integral- Derivative (PID) control was proposed. ASM adopted an assigning strategy based on dynamic critical paths to assign tasks to different process engines. Then, tasks in each engine were scheduled by a feedback strategy, which tuned parameters of a real-time tasks by elastic scheduling algorithm according to the output of admission controller and PID controller so as to maximize the processing capacity of the engine. Besides, to avoid instantaneous overloads, ASM adopted an admission control strategy, which controlled the number of tasks entering the system according to the current load of the system in order to ensure the stability of the system.

Key words: distributed workflow, adaptive scheduling, proportional-integral-derivative control, critical path

中图分类号: