隨著愈來愈多企業將他們的系統與服務部署到雲端平台上，雲端平台上計算資源的可靠性(reliability)以及可用性(availability)就更加重要。對於一個高可靠的雲端平台來說，容錯技術成為了一個系統中重要的部份。容錯技術讓雲端平台有了在硬體故障時可以持續提供服務的能力，即便後端計算資源的實體機器發生故障，也能通過冗餘，即備用的資源來接手運行的服務，使用者不會察覺到服務的中斷。 本研究基於中央大學平行與分散計算實驗室開發的NCU MFTVM容錯系統上繼續進行開發，讓原本未被容錯系統進行保護的虛擬機硬碟狀態，和其他狀態一起被同步至備援節點上。本研究針對了幾種虛擬機硬碟寫入狀況進行了實作上的修改，讓容錯系統在虛擬機對硬碟寫入情況下，提升應用程式在虛擬機上運作的效能，並縮短了虛擬機對外界回應的時間，提升了容錯系統的可靠性以及可用性。;More and more enterprise deploy their systems and services on Cloud platform, reliability and availability of computing resources on cloud platform will become increasingly important. For a highly reliable cloud platform, fault tolerance becomes an important part of the system. Fault-tolerant technology enables the cloud platform to continuously provide services in the event of hardware failure. Even if the physical machine of the computing resources fails, it can also take over the running services through redundancy, that is, standby resources, and users will not be aware of the interruption of the service. This research is based on the NCU MFTVM fault tolerant system developed by the Parallel and Distributed Computing Laboratory of National Central University. We implement a new feature so that the state of the virtual block device that is not protected by the origin fault tolerant system is synchronized to the Backup node together with other states. In this study, several disk workload of virtual machine are considered. We present several implementations that improve the performance of the application on the virtual machine, and reduce the response time of the virtual machine when the virtual machine writes to the hard disk. These improve the reliability and availability of the fault tolerance system.
