On Design and Implementation of Energy-aware and Fault Tolerant techniques for Cloud Computing Platforms
| Autor: | Ku, Chenhao, 古振浩 |
|---|---|
| Rok vydání: | 2013 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 101 Cloud computing utilizes virtualization techniques to build and deploy large number of virtual machines simultaneously, providing computing and storage services for anyone. However, the issue of energy consumption and increasing electricity price among other factors, the cost of maintaining huge cloud platform becomes increasingly higher. As result, methods and techniques for power saving in cloud platforms are of relevant importance. Additionally, in order to improve the stability of a huge cloud platform, it is necessary to embed fault tolerant technologies. Using such technologies, they can improve QoS of a cloud platform, yet reducing the maintenance and repair costs. This thesis provides a new monitoring system architecture which dynamically adjusts the frequency of monitoring according to the load of physical machines and the corresponding utilization of computing resources. As one physical machine is over loaded, monitoring system will increase the frequency of monitoring to improve the quality of monitoring system. On the other hand, as a physical machine is under loaded, monitoring system reduce the frequency of monitoring to save energy of computational resources. Still in this study, it is proposed a strategy to provide power saving and load balancing in a cloud platform. The algorithm in this strategy is based on the evaluation of loading from both physical machines and virtual machines. When physical machine is over utilized, physical machine migrate a part of virtual machine to reduce the loading of physical machine. On the contrary, when physical machine is under-utilized, all virtual machines in that host will be migrated to other physical machines, and our proposed strategy will send a signal to cloud platform’s controller to determine whether this host will be switch to sleep mode or not. With this approach, the cloud system will achieve the goals of both dynamic load balancing and energy saving. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|