基于蚁群算法的参数相关网格任务调度算法研究

任务调度是网格计算环境中影响系统性能的重要问题,目前大多数调度算法在调度时均把任务假定为元任务,而忽略任务间的约束关系。利用图论的思想将任务间依赖关系描述为满足一定条件的有向无环图（DAG）,通过DAG图可以精确描述任务的优先级,此外蚁群算法在元任务调度中能有效实现任务调度和负载均衡,因此在蚁群算法进行元任务调度结果的基础上,根据任务优先级对分配给资源的任务进行排序即得出了参数相关任务调度的结果。仿真结果显示,此方法在各种网格环境下均能解决相关任务的调度问题,而且保证系统的负载均衡。

Scheduling Interrelated Tasks in Grid Based on Ant Algorithm

Task scheduling is one of the most important factors of system performance in grid computing, many current algorithms which ignore the relations of tasks assume the scheduled task is Meta-task. According to Graph Theory, the interrelation of tasks can be described as Directed Acyclic Graphs, which can exactly define the priority of each task. In addition, the ant algorithm can effectively implement task scheduling and loading balance. The ultimate assign scheme was obtained by using the task priority to rearrange the Meta-task scheduling results generated by ant algorithm. Simulation demonstrates that the algorithm can schedule interrelated tasks in variety of grid environment and it also can guarantee loading balance of the system because of the using of ant algorithm.