【摘要】：計算中心的能耗問題越來越受到廣泛的關(guān)注，而在計算中心中主要使用虛擬化技術(shù)對計算資源進行管理和分配，因此，在虛擬化下的能耗性能平衡的研究是一個重要的研究方向。在虛擬化下，虛擬機管理器在分配系統(tǒng)資源中扮演著非常重要的角色，在傳統(tǒng)的虛擬CPU到物理CPU的映射中，由于虛擬CPU中運行的應(yīng)用類型是未知的，因此在傳統(tǒng)的調(diào)度中，并未考慮虛擬CPU的具體類型對于系統(tǒng)的能耗性能平衡造成的影響，使得在虛擬化下的系統(tǒng)能耗與性能平衡有著很明顯的提升空間。 本文以能耗延遲積作為能耗性能平衡的判定標準，通過對虛擬CPU的類型進行分類，將其分為CPU計算瓶頸型應(yīng)用，內(nèi)存瓶頸型應(yīng)用以及網(wǎng)絡(luò)I/O瓶頸型應(yīng)用對其分別進行討論，通過對資源競爭以及負載平衡的分析，針對虛擬CPU具體運行時狀態(tài)，得到了虛擬CPU到物理CPU映射方法，這個映射方法能夠提升系統(tǒng)的能耗與性能，在這個映射方法的基礎(chǔ)上，，進一步針對內(nèi)存瓶頸型應(yīng)用的特征，對傳統(tǒng)的頻率調(diào)節(jié)方法進行了改進， 通過在Xen平臺上的基準測試程序的實驗，結(jié)果表明，本文提供的虛擬CPU到物理CPU的映射方法和物理CPU頻率調(diào)節(jié)方法能夠在盡量保證系統(tǒng)性能的情況下有效的降低能耗，從而提升系統(tǒng)的能耗延遲積，達到一個更好的能耗性能平衡的狀態(tài)。
[Abstract]:More and more attention is paid to the problem of energy consumption in computing center, and virtualization technology is mainly used to manage and distribute computing resources in computing center. The study of energy performance balance in virtualization is an important research direction. In virtualization, the virtual machine manager plays a very important role in allocating system resources. In the traditional mapping of virtual CPU to physical CPU, because the type of application running in virtual CPU is unknown, it is in traditional scheduling. The effect of the specific type of virtual CPU on the energy consumption performance balance of the system is not considered, which makes the system energy consumption and performance balance in virtualization have obvious improvement space. By classifying the types of virtual CPU into CPU computing bottleneck applications, memory bottleneck applications and network I / O bottleneck applications, this paper takes energy delay product as the criterion of energy performance balance. Based on the analysis of resource competition and load balance, the mapping method of virtual CPU to physical CPU is obtained according to the specific runtime state of virtual CPU. This mapping method can improve the energy consumption and performance of the system. According to the characteristics of memory bottleneck application, the traditional frequency regulation method is further improved. The results show that, through the experiment of benchmark program on Xen platform, The mapping method of virtual CPU to physical CPU and the method of adjusting the frequency of physical CPU can effectively reduce the energy consumption while ensuring the system performance as far as possible, thus improving the energy consumption delay product of the system. Achieve a better energy performance balance state.
【學(xué)位授予單位】：上海交通大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TP391.9;TP393.05

【參考文獻】

相關(guān)期刊論文 前1條

1 葉可江;吳朝暉;姜曉紅;何欽銘;;虛擬化云計算平臺的能耗管理[J];計算機學(xué)報;2012年06期

本文編號：2196710