隨著硬體設備與網際網路頻寬不斷升級,在日常生活中以廣泛使用雲端運算,因此雲端運算已成為IT產業的熱門話題之一,在短短幾年中雲端運算已經融入我們的生活當中。而對於雲端運算平台,如何處理使用者請求虛擬化資源並做出適當的資源分配,將資源有效地進行分配,讓雲端環境能更有效地使用,儼然成為一個重要議題。隨著雲端運算的普及,開放原始碼的雲端運算平台紛紛出現,其中OpenStack受到許多科技大廠支持並投入開發,因此本研究針對OpenStack的資源分配模組Nova Scheduler中的FilterScheduler演算法進行研究與分析,觀察FilterScheduler演算法中Weighting階段僅考慮運算節點RAM的資源大小,實現的功能為運算節點剩餘的RAM越大為越適合放置的節點。因此本研究透過實驗虛擬機CPU、RAM與Disk資源對於運算節點的影響力,發現CPU與Disk對於虛擬機與運算節點有較大的影響力,而RAM的影響較小,因此本研究將FilterScheduler演算法中Weighting階段加入CPU核心數與Disk空間的判斷條件,增加OpenStack Nova服務的擴充性,使雲端環境資源能更有效地使用。最後透過本研究所設計的排程產生器,模擬使用者使用雲端環境,比較本研究所提出的演算法與原先FilterScheduler演算法之差異,實驗結果顯示本研究所提出的演算法加入CPU核心數與Disk空間判斷條件後,能有效提升雲端資源的使用,讓運算節點資源更平均分配。
Along with the continuous upgrade of hardware equipment and internet bandwidth, cloud computing has been widely used in daily life, therefore, cloud computing has become one of the hot topics in IT industry, in just a few years, cloud computing has got into our life completely. For cloud computing platform, how to handle the request of virtualized resource from the user and how to make appropriate resource allocation and create effective use of cloud environment has become an important topic. As cloud computing become more popular, cloud computing platform with open source code emerges one after another, among them, OpenStack has been supported by many enterprises and R&D has been launched, therefore, in this study, research and analysis will be conducted on FilterScheduler algorithm in Nova Scheduler, which was a resource allocation module of OpenStack. It was observed that in the Weighting stage of FilterScheduler algorithm, only the resource size of operation node RAM was considered, and the realized function was that the larger the remaining RAM in the operation node, more appropriate it was as a placement node, therefore, in this study, experiment was conducted to test the influence of CPU, RAM and Disk resources of virtual machine on the operation node, and it was found that CPU and Disk has stronger influence on the virtual machine and operation node, however, RAM has only less influence. Therefore, in this study, in the Weighting stage of FilterScheduler algorithm, number of CPU core and Disk space were added as judgment conditions, meanwhile, the expandability of OpenStack Nova service was added so that cloud environment resource can be more effectively used. Finally, through the schedule generator designed in this study, the cloud environment used by the user was simulated, meanwhile, the difference between the algorithm proposed in this study and the original FilterScheduler algorithm was compared. The experimental result showed that the algorithm proposed in this study, after adding the number of CPU core and Disk space judgment conditions, the use of cloud resource can be effectively enhanced, meanwhile, the operation node resources can be distributed more evenly.