Design And Implementation Of The Dual-Computer Hot-Standby System Based On Windows

Hot-standby system is an effective solution to improve the system availability. Based on the analysis of the factors that influence the availability, a research on all kinds of methods, Hot-standby system architecture, system fault diagnosis and checkpointing recovery is done. As the application of hot-standby system technology in critical transaction processing, hot-standby is a balance between availability and cost. A prototype of dual-computer hot-standby system is designed and implemented, and it can be extended to a multi-node High Availability (HA) cluster. To design a hot-standby system, architecture is the first issue to be solved. After the detailed analysis of architecture and storage technology of HA cluster, a Share Nothing architecture model based on a disk array is proposed. Complex cache consistency and distributed lock management (DLM) can be avoided in Share Nothing model. At the same time, the disk array ensures the high availability of shared data in the system. A Markov model is setup to analyze availability of this system.Successful failover ensure the high availability of a hot-standby system. failover action is divided into fault diagnosis and fault recovery. In several aspects of, such as, system node fault diagnosis, system service fault diagnosis and system resource fault diagnosis, an improved diagnosis technology based on multi monitor agents is implemented to achieve highly efficient diagnosis of system status. Hot-standby system is to decrease the number of single-point failures,and then to run continuously by using failover strategy. the failover process is composed of fault diagnosis and detection and fault recovery.Real Time File Mirrorring and CheckPointing and Rollback Recovery (CRR) is used to decrease the loss of fault then improve the availability. After the detailed analysis of the system architecture of Windows NT platform, a Real Time File Mirrorring system is designed and a checkpoint algorithm on WindowsNT platform is proposed and implemented.