近年來雲端運算興起,一躍成為當前最新議題,而數據中心提供商如何照使用者的服務層級協議(Service Level Agreement, SLA),用一有效決策來分配數據中心(Data Center)儼然成為一個挑戰。當然在資源管理中的配置排程也必須降低用電量,以提高數據中心提供商的利潤。在此我們定義了一個Utility函式,代表的是提供商的獲利滿意度。透過我們所提出的階層分析處理法(Analytical Hierarchical Process, AHP)來配置使用者到數據中心存取。本論文使用兩個演算法,Greedy和Max Fit,來模擬使用者虛擬機器(Virtual Machine, VM)需求的配置方法,以觀察配置上的電能耗費率,並且和一個傳統上分配使用者到距離最近的數據中心的方法比較。模擬結果發現,與傳統決策方法相比,使用AHP搭配兩演算法,Greedy及Max Fit,可以讓雲端服務供應商的滿意度明顯的提升。
In recent years, Cloud Computing has emerged as a hot topic in the world. Many cloud service providers have deployed data centers around the world to provide cloud services. In such globalized cloud services, it becomes very challenging to efficiently assign users to appropriate data center for services according to their SLAs (Service Level Agreements). The assignment decision must consider both user satisfaction and profit maximization of the service providers. In the thesis, we define a Utility function to represent the provider’s profit satisfaction, and present an AHP-based (Analytical Hierarchical Process) MADM (Multiple Attribute Decision Making) scheme to cope with the problem. Along with AHP-based decision making, we use Greedy and Max Fit algorithm for virtual machine allocation in each data center. Then we compare the performance of our AHP-based scheme with traditional distance-based assignment scheme. Simulation results illustrate that our scheme outperforms the distance-based scheme in both utility and profit of cloud service providers.