优化粒子群的云计算任务调度算法

Task Scheduling Algorithm of Cloud Computing Based on Particle Swarm Optimization

任务调度作为云计算的关键技术之一,却一直没有得到很好的解决。针对云任务调度的特点,基于基本粒子群优化(PSO)算法,文中提出了一种带极值扰动的相关性粒子群优化(EDCPSO)算法。该算法采用Copula函数去刻画随机因子间的相关结构,支持粒子合理利用自身经验信息和群体共享信息,解决了粒子群优化算法在寻优过程中没有考虑随机因子作用而造成全局优化能力不足的缺陷;采用添加极值扰动算子的策略,进一步改进粒子群优化算法,避免了粒子群优化算法在进化后期容易陷入局部寻优现象。仿真结果表明,在相同条件下,带极值扰动的相关性粒子群优化算法优于基本粒子群优化算法和Cloudsim原有调度算法,任务总的完成时间明显减少。

Howto schedule cloud tasks efficiently is one of the important issues to be resolved in cloud computing. The Extremum Disturbed Correlative Particle Swarm Optimization（ EDCPSO） algorithm based on basic Particle Swarm Optimization（ PSO） is proposed for the characteristics of cloud environment. It uses the Copular function to measure correlation structures among random factors,support of particles properly using the individual experience and social sharing information,resolving the demerit that the PSO algorithm lacks of the global optimization ability because of not considering the function of the random factors in the optimization process. Moreover,it uses the strategy of adding extremum disturbed arithmetic operators to improve further the PSO,which resolves the demerit of falling into local extremum in the late evolution for PSO. Simulation shows that EDCPSO is better than PSO and Cloudsim original scheduling algorithm in the same experiment conditions. That is to say,the algorithm can reduce the total completion time of tasks.