在雲端環境中，為使虛擬機器所提供之服務不中斷，故雲端環境需要虛擬機器具有高可用性(High Availability, HA)。若是因為機器故障而導致虛擬機無法繼續提供服務，對使用者來說每分鐘的故障代表著每分鐘的金錢損失。為降低提供服務之虛擬機中斷時間。一般而言，當虛擬機器錯誤發生時，雲端系統必須要很快地偵測出錯誤，然後進行快速復原。在雲端HA的作法上，當檢測出運算節點發生故障，就需要將虛擬機疏散(Evacuate)至其他運算節點。本研究針對的是基於OpenStack雲端平台之各種現存虛擬機器疏散策略的比較，了解最迅速的排程法將虛擬機更快的重啟，達到虛擬機快速復原，以提高雲端系統之高可用性。藉由本論文的實驗發現，最佳情況下 RR較 Random多減少了 約 10%的時間，較 Nova Filter Scheduler減少了 減少了 20%的;In the cloud environment, to avoid the services provided by the virtual machine being interrupted, the cloud environment requires virtual machine with high availability (HA). If the suspension of virtual machine leading to stop providing services due to the mechanical failures, failures per minutes will turn into cost for users. In order to reduce the interruption time of virtual machine, when a virtual machine error occurs, the cloud systems should detect and recover errors as soon as possible. In a real situation of HA in the cloud, when the computing node failure is detected, it is necessary to evacuate the virtual machine to other computing nodes. This study is aimed at analyzing the difference between the existing virtual machines evacuation strategies based on OpenStack cloud platform, to acknowledge the most rapid method to schedule restarting the virtual machine faster and to achieve rapid recovery to improve the availability of cloud systems. According to the experiments in this paper, RR reduced about 10% of the time less than Random, and about 20% of the time less than Nova Filter Scheduler.
