【作者】 王晓丽；

【导师】 王宇平；

【作者基本信息】 西安电子科技大学 ， 计算机应用技术， 2014， 博士

【摘要】 随着云计算数据中心规模的不断扩大，能源消耗和通信带宽受限问题已经成为制约其发展的主要瓶颈之一。对于前者，提高服务器的能源利用率是降低数据中心能耗的有效途径。如何通过恰当的调度策略来提高服务器的能源利用率是本文的主要研究内容之一。对于后者，由于数据中心的带宽有限，是十分宝贵的资源，如何通过恰当的调度策略来提高任务的数据本地化执行率，是节省数据中心带宽使用量的关键，也是本文的另一主要研究内容。本文的主要创新点可以概括为：通过恰当的调度策略解决云计算目前最为突出的能耗问题和带宽问题。即通过有效的调度策略调整服务器上的数据部署和任务分配，达到最大化服务器能源利用率和任务执行的数据本地化率，从而降低数据中心的总能耗并减少带宽的使用量。基于此思想，我们为云环境下的能耗与带宽问题先后建立了三组优化模型，并设计了相应的求解算法。具体工作如下：1.建立了节能大规模任务调度模型，并设计了一种全局优化遗传算法求解该模型。通过合理的任务调度策略调整服务器的CPU利用率，从而提高服务器的能源利用率。同时，该模型通过确保任务100%的数据本地化执行率，提高任务的执行效率、节省数据中心的带宽使用。2.建立了存储与计算融合的节能大规模优化模型，并设计了一种双层遗传算法求解该模型。将数据部署与任务调度相结合，通过调整服务器的资源利用率（CPU和硬盘利用率），提高服务器的能源利用率，同时，该模型通过确保任务100%的数据本地化执行率，提高任务的执行效率、节省数据中心的带宽使用。3.建立了存储与计算融合的多目标大规模优化模型，并基于MOEA/D设计了一种多目标双层遗传算法求解该模型。为决策者提供一组可选的数据部署策略和任务调度方案，使得在满足数据中心当前带宽需求的条件下，获得最高的能源利用率。4.由于云环境下的任务规模往往成千上万，因此所建立的调度模型均为大规模优化模型。通过在遗传算法中引入局部搜索算子，提高了模型的求解效率、加快算法的收敛速度。实验表明，所建立的三组优化模型是合理的，对应的模型求解算法能有效的提高数据中心的能源利用率，并降低数据中心的带宽使用量。更多还原

【Abstract】 With the constant expansion of data centers, huge energy consumption and limited band-width have become one of the main factors restricting the development of cloud computing. Forthe former, improving servers’ energy efficiency is one of the effective ways to reduce energyconsumption of data centers. Thus one of the main issue of this paper is to study how to im-prove servers’ energy efficiency through appropriate scheduling strategies. For the latter, sincebandwidth is a very limited and valuable resource in a data center, how to save bandwidth byreasonable scheduling strategies is another main issue of this paper.The main objective of this thesis is to design appropriate scheduling strategies to solve twoof the most prominent issues of cloud computing—huge energy consumption and limited band-width. That is to say, adjusting data deployment and task allocation among servers by effectivescheduling strategies, so as to maximize the energy efficiency of servers and the data localityratio of tasks. Thereby, the total energy consumption and bandwidth usage could be reduced.Based on these, we built the three sets of optimization models and design their correspondingalgorithms for the energy consumption and bandwidth problems under cloud computing. Themain contributions are as follows:1. Build an energy-aware large-scale task-scheduling model and design a global optimiza-tion genetic algorithm to solve the model. The proposed model adjust the CPU utilizationof servers through reasonable task-scheduling strategies, so as to improve the energy effi-ciency of servers. Meanwhile, by ensuring100%data locality ratio of tasks, it improvesthe execution efficiency of tasks and saves the bandwidth usage of data centers.2. Build a data deployment and task allocation combined energy-aware large-scale modeland design a bi-level genetic algorithm to solve the model. The proposed model adjustthe resource utilizations of servers, including the CPU and Disk utilization, by combiningdata deployment policies and task-scheduling strategies, so as to improve the energy effi-ciency of servers. Meanwhile, by ensuring100%data locality ratio of tasks, it improvesthe execution efficiency of tasks and saves the bandwidth usage of data centers.3. Build a data deployment and task allocation combined multi-objective model and designa multi-objective genetic algorithm based on MOEA/D to solve the model. The proposedmodel provide a set of optional data deployment policies and task-scheduling strategiesto the decision makers, so as to get the maximum energy efficiency of servers under theprerequisite of meeting the current demand for bandwidth.4. As tasks involved in cloud computing are usually tens of thousands, the proposed mod-els are all large-scale optimization ones. In order to speed up the solving process of the proposed models and accelerate convergent speed of the proposed algorithm, local searchoperators is introduced in genetic algorithms. Finally, numerical experiments are madeand the results indicate that the proposed models are reasonable and their correspondingalgorithms can effectively improve the energy efficiency of servers and save the band-width usage of data centers.更多还原