云计算环境下基于装箱策略的多目标调度方法

项目名称： 云计算环境下基于装箱策略的多目标调度方法

项目编号： No.61472139

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 罗飞

作者单位： 华东理工大学

项目金额： 80万元

中文摘要： 云计算环境下的任务调度问题是一个NP完全问题，且存在着调度目标少、调度模型效率低、调度层次单一、算法性能评价方法缺失等亟待解决的问题。针对这些问题，提出基于装箱策略的多目标调度机制，内容包括：（1）针对最小化应用完成时间、减少云平台能耗、提高系统的健壮性等多目标，定义云计算 环境下应用-资源匹配关系，进而研究云计算环境下多目标调度策略的一般数学模型；（2）综合考虑云计算节点资源与计算任务的不确定性，以层次化思想分别研究以应用为中心、以资源为中心和以效益为中心的资源模型；（3）为了应对各种云计算应用环境，基于装箱策略，研究利用批量装箱机制、变尺寸装箱机制和Open-end装箱机制设计多目标调度算法；（4）深入研究算法的调控机制，提供一种针对启发式算法的评价模型和方法。项目的实施为云计算中的多目标任务调度提供一种可行的方法实践，进而提升云计算平台的综合性能，拓展云计算应用的范围。

中文关键词： 云计算；任务调度；多目标；装箱；性能优化

英文摘要： The task scheduling is a NP-complete problem in the cloud computing environment, and there are some problems needed to be solved, such as scheduling with insufficient objects, the low efficiency of the scheduling model, designing the scheduling policy in a single system level, and deficiency of the way to evaluate the performance of algorithms. Aiming at these problems, a multi-objective scheduling strategy based on the bin-packing policy is proposed, and the research contents are presented as follows. First, the scheduling objects include minimizing application's makespan, reducing the energy consumption of the platform and improving the system's robustness, so that mapping relationship between the application and resources are defined, and a general mathematical model for the multiple scheduling objects in the cloud is presented. Second, considering the uncertainty of the computing resources and tasks in the cloud, a hierarchical resource model is discussed in the application level, in the resource level, as well as in the benefit level. Third, in order to be adaptable to different cloud environments, several multi-objective scheduling algorithms based on bin packing strategy are proposed, such as the scheduling algorithm with batch packing, the one with variable-size bin packing, as well as the with open-end bin packing. Finally, with a thorough study on the control of the algorithms, a framework to evaluate the heuristic algorithms is provided. The implementation of the project presents a feasible method for multi-objective scheduling in the cloud environment, which will enhance the overall performance of the cloud computing platform, and expand the range of applications in the cloud.

英文关键词： Cloud Computing;Task Scheduling;Multi-objective;bin-packing;Performance optimization