基于 DPSO 的云工作流任务自适应调度策略

摘要

云工作流系统中的任务调度问题属于典型的NP难题，同时由于计算资源异构性、复杂性及用户需求的动态性可能导致系统过载。为了解决或避免此类问题的发生，本文提出一种带动态反馈机制的任务自适应分配方法，并结合离散粒子群优化算法（ Discrete Particle Swarm Optimization ，DPSO），利用任务预测执行时间模型来优化任务分配方案。仿真实验表明该方法可保证系统负载平衡，当任务数大于150时能够使任务调度时间最短。%The task scheduling in cloud workflow system is a typical NP-complete problem , and the heterogeneity of computing resource and the complexity and dynamics of the requirement may lead to system overloads .We present a self-adaptive task scheduling model which adops Discrete Particle Swarm Optimization algorithm to schedule tasks to virtual machines .Simulation results show that the proposed strategy guarantees a good load balance and achieves shorter scheduling time when the number of tasks is more than 150.