Research On The Optimization Mechanisms For Hot Standby Based On Remus

Virtualization technology, the key technology of cloud computing, could enhance thesecurity of operating system through creating multiple isolated virtual machines in aphysical machine. Besides, Virtual machine dual hot backup technology is a high efficiency,low cost fault-tolerant mechanism with the help of the characteristics of virtual machines,for example cross-platform portability, detached from the underlying hardware and so on.To achieve state synchronization and data synchronization between host node andbackup node, traditional hot standby technology constantly generate checkpoint, sendreal-time status updates of host virtual machine to backup virtual machine. However,traditional checkpoint technology will cause bandwidth delay which will affect the highavailability of this technology when it encountered intensive applications during backup,such as Remus based on Xen architecture. In addition, during the process of backupheartbeat may lose effectiveness which led to false positives about virtual machine’s stateand infect system’s normal operation. Therefore, this paper proposed two optimizationsbased on Remus hot standby mode. One is to compress the memory checkpoint, the other isto design a client-oriented hot standby mechanism.Memory checkpoint compression scheme introduced checkpoint compression basedon XOR-RLE algorithm to reduce the bandwidth-delay. Geared to the needs of the client’shot standby system retains the Remus heartbeat mechanism and the state of main virtualmachine can be judged for whether the backup machine should take over the control ofexternal service with the lord for heartbeat failure between the virtual machine. Thementioned mechanism has eased the erroneous judgement of the main virtual machine,which has improved the stability of the system.Through comparison of bandwidth-delay before and after the improvement,checkpoint compression based on XOR-RLE algorithm effectively reduces the bandwidthdelay caused by the intensive applications, and the client-oriented hot standby mechanismsubstantially reduced false positives during hot standby process.