中文摘要：

云计算系统的高效能调度优化是当前重要的研究课题,面向异构云环境的多目标优化调度方法研究具有重要意义.云计算环境下的能耗和性能优化管理是NP-HARD的多目标组合优化问题,目前一般启发式调度系统大多采用带约束的性能或能耗的单目标优化计算方法,不能完全满足复杂云计算系统资源约束动态性与管理需求多样性的需求.基于传统进化优化的随机搜索算法应用于云环境下的DAG任务的多目标调度优化,计算开销大、计算实时性不足,文中提出了新的Memetic优化方法以解决异构云环境多目标调度优化问题.首先,文中针对异构云环境多目标调度优化问题,构建了一般性的数学定义;其次,针对该问题设计了多目标Memetic优化算法,采用基于解结构相关信息的Memetic局部搜索算子加速调度方案的局部优化能力,以提高算法的收敛速度、降低计算开销.实验结果表明,应用所提出的多目标Memetic优化算法进行异构云环境能耗和性能多目标调度优化,比传统方法具有更好的计算效率、解集多样性与收敛性能.

英文摘要：

Highly efficient scheduling optimization of cloud computing system is an important research subject, and the multi-objective optimal scheduling algorithm on the heterogeneous cloud is meaningful. The problem of energy consumption and performance optimization management on cloud is NP-hard multi-objective combinatorial optimization problem. Currently, most of general heuristic based scheduling algorithm adopt single objective optimization method which binding performance or energy consumption calculation. These methods cannot fully satisfy the dynamics constraints of complex cloud computing system resources, and the diversity of management. The computation overhead of general stochastic search algorithm based on evolutionary optimization for DAG scheduling on cloud is expensive, and real-time calculation is insufficient. For these reasons, a new Memetic optimal algorithm is proposed. In this paper, we define the problem of multi-objective scheduling optimization on the heterogeneous cloud. And then, a Multi-objective Memetic algorithm is proposed, which use memetic local search technique based on the related information of solution structure to improve the local optimization ability, this technique could improve the algorithm convergence speed and reduce the computational overhead of algorithm.The experiment results show that the proposed method has better computation efficiency, diversity of solution set and convergence performance than traditional methods.