摘要： 目前研究的动态任务调度算法都基于集中式或部分分布式网格系统，系统中心节点(组)进行资源管理。该文提出一种面向无资源管理的完全分布式网格系统动态任务调度算法DE Scheduling。该算法使用任务冗余调度算法屏蔽解决系统的动态性问题，通过动态调节冗余量减少无效计算和保证系统负载均衡。使用给定平均连接度的无标度网络演化模型构造具有1 000个节点的Internet网络模型仿真任务处理过程。仿真结果表明，任务数为10 000~100 000时该系统冗余调度次数均为2次，冗余计算量占总计算量的比例不超过0.35%，且随着任务数增加而递减。

关键词: 分布式网格系统, 任务调度, 冗余调度

Abstract: A dynamic scheduling algorithms studied in references are based on centralized grid system or part-distributed grid system, in which there are center node(s) to manage resources. A dynamic scheduling algorithm for the fully distributed grid system without resource management, named DE Scheduling, is proposed. In the algorithm, the redundant scheduling is used to solve the dynamic of environment, and the time of redundant scheduling is set dynamically to reduce void calculations and balance the load of the system. A group of simulations have been done by using the Internet model with 1 000 nodes based on the evolving model for scale-free network with given mean connected degree. When the number of tasks is from 10 000 to 100 000, the time of redundant scheduling always equals 2, the redundant calculations account of the total amount is less than 0.35%, and reduces with the increasing number of tasks.

Key words: distributed grid system, task scheduling, redundant scheduling

中图分类号: 

• TP312