基于混合粒子群分布估计算法的Hadoop任务调度优化策略

HADOOP TASK SCHEDULING OPTIMISATION STRATEGY BASED ON HYBRID PARTICLE SWARM OPTIMISATION-ESTIMATION OF DISTRIBUTION ALGORITHM

在一个异构的网格环境下,Hadoop异构任务调度的目的是有效地利用资源和共享可用的资源之间的负载,这样的任务调度问题是NP-Hard问题。提出一种基于混合粒子群分布估计算法(HPSO-EDA)的任务分配策略。新的HPSO-EDA引入分布估计算法的建立概率模型和随机抽样操作来替代速度和位置的更新操作来引导最优解的进化,提高算法的收敛速度,防止算法陷入局部最优化解。通过实验仿真表明:HPSO-EDA比传统PSO和EDA能在更短的时间里产生更好的结果。

In a heterogeneous grid environment, the ~urpose of Hadoop heterogeneous task scheduling is to effectively use resources and to share the load among available resources, such task scheduling problem is the NP-hard problem. This paper proposes a task allocation strategy which is based on hybrid particle swarm optimisation-estimation of distribution algorithm （HPSO-EDA）. The new HPSO-EDA introduces the probability model building and random sampling operation of EDA to replace the velocity and position update operations to guide the evolution of optimal solution, it is in order to improve the convergence rate of the algorithm and prevent the algorithm falling into local optimal solution. It is demonstrated by the experimental simulation that the HPSO-EDA algorithm proposed in the paper can generate better results in shorter time than the traditional PSO and EDA algorithms.