云计算平台中的能耗管理方法

发布时间：2018-04-10 14:20

本文选题：云计算 + 能耗模型　；参考：《南京邮电大学》2013年硕士论文

【摘要】：高能耗是云计算系统的一个主要问题，且随着近年来云计算规模的日益扩大，其能耗开销也愈加严重。本文针对云计算系统中的空闲和奢侈能耗，以动态电源管理及资源调度为研究内容，以节能为主要研究目标，主要作了以下四个方面的工作：（1）总结了国内外节能技术的研究现状，重点阐述了云计算的相关技术、现有的动态电源管理策略及云环境下的资源调度算法。（2）本文充分考虑资源在休眠、空闲、工作和转换等不同状态下的多种能耗开销，提出一种基于不同状态的能耗估算模型（Energy Consumption Estimation Model based on DifferentStates，ECEMDS），并且用多功能计量插座对其进行验证。（3）针对云计算系统中空闲能耗，本文基于指数平均算法，提出了一种自适应的空闲时间预测策略。该策略引入了自适应权值调节因子，动态调节历史空闲时间对预测空闲时间的影响，同时结合分段滑动窗口的思想，将滑动窗口内的空闲时间分为长、中、短三类，，取每类个数最多的空闲时间的平均值作为指数平均算法的实际空闲时间值。该策略通过对下一段空闲时间的预测，决定是否要切换物理主机的状态以降低空闲能耗。实验结果说明本文提出的策略预测准确率较高，系统响应时间少，节能效果好。（4）本文针对奢侈能耗，提出一种云环境下节能的资源调度算法。首先，对云环境中的资源调度进行建模。接着，提出一种基于改进Min-Min算法的最小能耗资源调度算法Min-Energy，在满足云任务QoS需求的基础上，以云环境下每个任务的能耗最小为目标调度资源。算法按照优先级对任务排序，然后估算每个任务在各个资源上的能耗总和，选择每个任务的最小能耗对应的资源进行调度。在CloudSim平台的仿真结果表明，Min-Energy算法在完成时间和能量消耗方面均具有较好的性能，能够达到节能的目的。
[Abstract]:High energy consumption is a major problem in cloud computing systems, and with the increasing scale of cloud computing in recent years, its energy consumption is becoming more and more serious.Aiming at the idle and extravagant energy consumption in cloud computing system, this paper takes dynamic power management and resource scheduling as the research content, and takes energy conservation as the main research goal, mainly doing the following four aspects of work:This paper summarizes the research status of energy saving technology at home and abroad, and focuses on the related technologies of cloud computing, the existing dynamic power management strategy and resource scheduling algorithm in cloud environment.In this paper, energy Consumption Estimation Model based on difference states is proposed and verified by multifunctional metering sockets, considering the energy cost of resources in different states, such as dormancy, idle, work and conversion. The energy consumption estimation model based on different states is presented in this paper.Aiming at the idle energy consumption in cloud computing system, an adaptive idle time prediction strategy based on exponential average algorithm is proposed in this paper.The strategy introduces the adaptive weight adjustment factor, dynamically adjusts the influence of the historical idle time on the prediction of idle time, and combines the idea of piecewise sliding window, divides the idle time in the sliding window into three categories: long, medium and short.The average of the idle time with the largest number of each class is taken as the actual idle time value of the exponential average algorithm.By predicting the next idle time, the strategy determines whether to switch the state of the physical host to reduce idle energy consumption.The experimental results show that the strategy proposed in this paper has higher prediction accuracy, less system response time and better energy saving effect.In this paper, a resource scheduling algorithm for energy saving in cloud environment is proposed for luxury energy consumption.Firstly, resource scheduling in cloud environment is modeled.Then, a Min-Energy scheduling algorithm based on improved Min-Min algorithm is proposed. On the basis of satisfying the QoS requirements of cloud tasks, the minimum energy consumption of each task in the cloud environment is taken as the target.The algorithm sorts the tasks according to the priority, then estimates the total energy consumption of each task on each resource, and selects the resources corresponding to the minimum energy consumption of each task to schedule.The simulation results on the CloudSim platform show that the Min-Energy algorithm has better performance in terms of completion time and energy consumption, and can achieve the purpose of energy saving.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP3

【参考文献】