满足帕累托最优的多目标云工作流调度算法

为了同步考虑用户的任务QoS需求和云资源提供方的收益,提出一种云环境中满足帕累托最优的多目标最优化DAG(Directed Acyclic Graph)粒子群算法MODPSO(Multi-objective DAG Particle Swarm Optimization)。综合考虑任务执行跨度、执行代价与执行能耗的三目标同步最优化,设计基于DVFS的离散PSO调度优化方法。重新定义PSO的种群粒子进化过程和更新规则,进而得到多目标优化工作流调度解。通过人工合成工作流和现实科学工作流进行仿真测试,并对算法性能进行分析。结果表明,该算法可以通过非支配集的方式实现冲突多目标的调度优化求解。在满足用户QoS的同时,得到最优解的Pareto边界集,实现调度性能与系统能耗的均衡。

MULTI-OBJECTIVE CLOUD WORKFLOW SCHEDULING ALGORITHM SATISFYING PARETO OPTIMALITY

In order to synchronously consider the user's task QoS requirements and the benefits of cloud resource providers,we proposed a multi-objective DAG particle swarm optimization(MODPSO)to satisfy Pareto optimality in cloud environment.Considering the three-objective synchronization optimization of task execution span,execution cost and execution energy consumption,we designed a discrete PSO scheduling optimization based on DVFS.The evolutionary process and updating rules of PSO were redefined,and then the multi-objective optimal workflow scheduling solution was obtained.By the simulation experiments on synthetic workflow and real scientific workflow,we analyzed the performance of the algorithm.The results show that the algorithm can achieve multi-objective scheduling optimization by non-dominant set.While satisfying the user's QoS,the Pareto boundary set of the optimal solution can be obtained to achieve the balance between scheduling performance and system energy consumption.