一种可靠性驱动的云工作流调度遗传算法

Cloud workflow scheduling genetic algorithm satisfying reliability driven

为了解决云环境中工作流调度的可靠性问题,提出了一种基于可靠性驱动信誉度模型的工作流调度遗传算法RDR-GA。算法以工作流执行跨度makespan与可靠性最优化为目标,设计了一种基于时间依赖的可靠性驱动信誉度模型,通过该模型可以有效评估资源可靠性。同时,为了寻找遗传最优解,算法设计了新的遗传进化和评估机制,包括:以进化算子对调度解中的任务—资源映射进行遗传进化;以两阶段MAX-MIN策略评估并决定调度解的任务执行序列。仿真实验结果表明,满足可靠性驱动的信誉度算法不仅能够以更精确的信誉度改善工作流应用执行可靠性,而且能够以比同类遗传算法更快的收敛速度得到进化更优解。

In order to solve the reliability problem of workflow scheduling in cloud computing environment,this paper proposed a workflow scheduling genetic algorithm based on reliability-driven reputation model.This algorithm defined the makespan and reliability of workflow scheduling as the optimization objectives,and designed a reliability-driven reputation model based on time dependent.It could effectively evaluate the reliability of a resource by this model.At the same time,to find genetic optimal solution,this algorithm designed a novel evoluation and evaluation mechanism,included:the evolution operators evolved the task-resource mapping of a scheduling solution and the evaluation step determined the task order of solutions by using the two-phase MAX-MIN strategy.The simulation experimental results show that the satisfying reliability-driven reputation algorithm not only can improve the reliability of an workflow application execution with more accurate reputations,but also can provide and evolve to better solutions with a faster convergence speed than another genetic algorithm.