自适应动态调整粒子群的云计算任务调度

ADAPTIVE DYNAMIC ADJUSTMENT OF PARTICLE SWARM OPTIMIZATION FOR CLOUD COMPUTING TASK SCHEDULING

为了高效地实现云计算任务调度,融合改进的分数阶达尔文粒子群算法和多目标函数构造,提出一种新的云计算任务调度算法。对分数阶达尔文粒子群算法进行全方位改进,基于粒子群适应度动态调整惯性权重系数以自适应搜索最优解;利用粒子自身进化信息定义进化因子,结合进化因子并利用高斯图函数调整分数阶次α系数以实现快速收敛;借助Levy飞行随机扰动对局部最优位置进行位置扰动以提高跳出局部最优的能力;综合最短等待时间、资源负载均衡程度及任务完成所耗费用等三个目标构造任务调度满意度函数,以此搜索任务调度最优解。仿真实验表明,与其他粒子优化算法相比,该算法有较快的收敛速度和较高的寻优精度;在任务调度中,该算法与其他三种调度算法相比,在较低的截止时间未完成率下实现了虚拟资源的均衡负载。

In order to efficiently implement cloud-computing task scheduling, the improved fractional Darwin particle swarm optimization and multi-objective function construction were combined, and a new cloud computing task-scheduling algorithm was proposed. The algorithm improved the fractional-order Darwin particle swarm optimization. Based on the particle swarm fitness, the inertia weight coefficient was dynamically adjusted to adaptively search the optimal solution. Evolutionary factors were defined based on their own evolutionary information,and the fractional α coefficients were adjusted using evolutionary factors and Gauss graph functions to achieve fast convergence. Levy flight stochastic disturbance was used to disturb the locally optimal position to improve the ability of jumping out of the locally optimal position. Task scheduling satisfaction function was constructed by synthesizing the shortest waiting time, load balancing degree of resources and the cost of task completion to search the optimal solution of task scheduling. Simulation results show that compared with other particle optimization, this algorithm has faster convergence speed and higher optimization accuracy. In task scheduling, compared with the other three scheduling algorithms, the improved scheduling algorithm achieves a balanced load of virtual resources under a lower cut-off time unfinished rate.