面向节能的云计算任务调度策略研究

发布时间：2018-03-08 16:50

本文选题：绿色计算　切入点：云计算　出处：《哈尔滨工业大学》2013年硕士论文　论文类型：学位论文

【摘要】：随着信息技术的高速发展，信息技术行业所带来的能量消耗也成为人们十分关注的问题之一。随着云计算的发展逐渐成熟，相关的应用正在逐年增加，由于云计算数据中心服务器及其配套设备规模的高速增长，快速攀升的能耗已成为影响企业利润的重要因素，研究如何对数据中心的资源和任务进行优化管理，以降低能耗、减少污染对企业和环境保护都有着重要的意义。云计算数据中心通常包含一个服务器机群，这些服务器同时运行大量的应用程序，这种情况就可以对数据中心的应用负载进行整合，用较小数量的服务器运行任务，使服务器的各项资源都能得到充分的利用的同时又不会出现资源争用的情况，从而达到降低成本，节约能耗的目的，这就是本文所要研究的内容。由于不同任务对CPU、内存等各种计算资源的需求量不同，为了使数据中心服务器各项资源得到充分利用，首先需要对任务对不同计算资源的需求量进行预测，针对这一问题，本文首先提出了基于神经网络的程序资源消耗预测模型，使用这一预测模型对云计算任务各项计算资源消耗进行预测，该模型以影响程序运行资源消耗的各项因素作为神经网络输入，以程序运行所消耗的时间、CPU利用率、内存使用量、硬盘使用量作为网络输出，，收集程序运行的历史数据作为神经网络的训练和测试样本，实现对程序性能及资源使用的预测。根据云计算任务各项资源消耗量的预测结果，对数据中心的任务和服务器各项资源进行整合，优化任务调度方案。为了减少运行主机并使其各项硬件资源得到充分的利用，同时又能够避免资源争用的情况出现，本文将任务分配问题转化为一个多维多背包问题进行求解，由于任务分配问题是一个NP完全问题，本文设计采用混合遗传算法对该问题求解，以能耗最小作为目标函数，求得任务分配问题最低能耗的优化解，从而实现降低能耗，节约成本的目的。
[Abstract]:With the rapid development of information technology, the energy consumption brought by the information technology industry has become one of the problems that people pay close attention to. With the development of cloud computing, the related applications are increasing year by year. Due to the rapid growth of cloud computing data center servers and their supporting equipment scale, the rapidly rising energy consumption has become an important factor affecting the profits of enterprises. This paper studies how to optimize the management of data center resources and tasks in order to reduce energy consumption. Reducing pollution is of great significance to enterprises and environmental protection. Cloud computing data centers typically contain a cluster of servers that run a large number of applications at the same time, so that the application load of the data center can be consolidated to run tasks with a smaller number of servers. So that all the resources of the server can be fully utilized without the situation of resource contention, so as to achieve the purpose of reducing cost and saving energy consumption, this is the content of this paper. Because different tasks require different computing resources, such as CPU, memory and so on, in order to make full use of the resources of the data center server, it is necessary to forecast the demand of different computing resources for different tasks, aiming at this problem. In this paper, a program resource consumption prediction model based on neural network is proposed, which is used to predict the computing resource consumption of cloud computing tasks. In this model, the factors that affect the consumption of running resources are taken as the input of neural network, and the CPU utilization, memory usage and hard disk usage are used as the network output. The historical data of program running are collected as training and test samples of neural network to predict program performance and resource usage. According to the forecast results of resource consumption of cloud computing task, the task of data center and the resource of server are integrated, and the task scheduling scheme is optimized. In order to reduce the running host and make full use of its hardware resources, At the same time, the problem of resource contention can be avoided. In this paper, the task assignment problem is transformed into a multi-dimensional multi-knapsack problem, because the task assignment problem is a NP-complete problem. In this paper, a hybrid genetic algorithm is used to solve the problem. With the minimum energy consumption as the objective function, the optimal solution of the minimum energy consumption of the task assignment problem is obtained, so as to reduce the energy consumption and save the cost.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP308;TP301.6

【参考文献】