近年來雲端計算蓬勃發展，IaaS類型的雲端應用需要更多機器來提供服務，所以能將一部實體機器當多部虛擬機器的虛擬化技術成為相當熱門的議題。虛擬化技術能提升實體機器的使用率，但也提升機器故障所造成的損失，高可用性成為相當有效的解決方案，高可用性在故障發生時仍然會有一段停機時間及遺失部分資訊，所以容錯成為更有效的解決方案。容錯系統依照不同特性及運作方式區分為數種架構，持續同步是藉由不中斷的在主要機器與備援機器間同步虛擬機器狀態達成容錯系統中狀態一致的基本要求，也是本研究所採用的架構。本研究基於QEMU-3.0.0設計及實作基於持續同步的容錯系統，除了引入其他虛擬機研究的相關技術來改進本研究，本研究還藉由追蹤及分析虛擬機實作的程式碼來設法改進容錯系統。;Infrastructure as a Service (IaaS) on the cloud requires more machines to provide services. Therefore the virtualization technology that can use a physical machine as multiple independent virtual machines has become a hot topic. Virtualization technology can increase the utilization rate of physical machines, but it also increases the losses caused by machine failures. Fault Tolerance (FT) becomes a very effective solution for this problem. A fault-tolerant VM system can use several architectures and technology for implementation. Continuous checkpointing is one popular way for achieving the state of the fault-tolerant system by uninterrupted synchronization of the virtual machine states between the main machine and the backup machine, and is also the architecture adopted in this study. This study is based on “QEMU-3.0.0”. We present several implementation techniques and demonstrate its efficiency and overhead in the paper.
